3RNN/Lib/site-packages/tensorflow/python/ops/gen_image_ops.py
2024-05-26 19:49:15 +02:00

4879 lines
231 KiB
Python

"""Python wrappers around TensorFlow ops.
This file is MACHINE GENERATED! Do not edit.
"""
import collections
from tensorflow.python import pywrap_tfe as pywrap_tfe
from tensorflow.python.eager import context as _context
from tensorflow.python.eager import core as _core
from tensorflow.python.eager import execute as _execute
from tensorflow.python.framework import dtypes as _dtypes
from tensorflow.security.fuzzing.py import annotation_types as _atypes
from tensorflow.python.framework import op_def_registry as _op_def_registry
from tensorflow.python.framework import ops as _ops
from tensorflow.python.framework import op_def_library as _op_def_library
from tensorflow.python.util.deprecation import deprecated_endpoints
from tensorflow.python.util import dispatch as _dispatch
from tensorflow.python.util.tf_export import tf_export
from typing import TypeVar, List, Any
from typing_extensions import Annotated
TV_AdjustContrast_T = TypeVar("TV_AdjustContrast_T", _atypes.Float32, _atypes.Float64, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt8)
def adjust_contrast(images: Annotated[Any, TV_AdjustContrast_T], contrast_factor: Annotated[Any, _atypes.Float32], min_value: Annotated[Any, _atypes.Float32], max_value: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, _atypes.Float32]:
r"""Deprecated. Disallowed in GraphDef version >= 2.
Args:
images: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int16`, `int32`, `int64`, `float32`, `float64`.
contrast_factor: A `Tensor` of type `float32`.
min_value: A `Tensor` of type `float32`.
max_value: A `Tensor` of type `float32`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "AdjustContrast", name, images, contrast_factor, min_value,
max_value)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return adjust_contrast_eager_fallback(
images, contrast_factor, min_value, max_value, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"AdjustContrast", images=images, contrast_factor=contrast_factor,
min_value=min_value, max_value=max_value, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"AdjustContrast", _inputs_flat, _attrs, _result)
_result, = _result
return _result
AdjustContrast = tf_export("raw_ops.AdjustContrast")(_ops.to_raw_op(adjust_contrast))
def adjust_contrast_eager_fallback(images: Annotated[Any, TV_AdjustContrast_T], contrast_factor: Annotated[Any, _atypes.Float32], min_value: Annotated[Any, _atypes.Float32], max_value: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, _atypes.Float32]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, _dtypes.float32, _dtypes.float64, ])
contrast_factor = _ops.convert_to_tensor(contrast_factor, _dtypes.float32)
min_value = _ops.convert_to_tensor(min_value, _dtypes.float32)
max_value = _ops.convert_to_tensor(max_value, _dtypes.float32)
_inputs_flat = [images, contrast_factor, min_value, max_value]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"AdjustContrast", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"AdjustContrast", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_AdjustContrastv2_T = TypeVar("TV_AdjustContrastv2_T", _atypes.Float32, _atypes.Half)
def adjust_contrastv2(images: Annotated[Any, TV_AdjustContrastv2_T], contrast_factor: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, TV_AdjustContrastv2_T]:
r"""Adjust the contrast of one or more images.
`images` is a tensor of at least 3 dimensions. The last 3 dimensions are
interpreted as `[height, width, channels]`. The other dimensions only
represent a collection of images, such as `[batch, height, width, channels].`
Contrast is adjusted independently for each channel of each image.
For each channel, the Op first computes the mean of the image pixels in the
channel and then adjusts each component of each pixel to
`(x - mean) * contrast_factor + mean`.
Args:
images: A `Tensor`. Must be one of the following types: `half`, `float32`.
Images to adjust. At least 3-D.
contrast_factor: A `Tensor` of type `float32`.
A float multiplier for adjusting contrast.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "AdjustContrastv2", name, images, contrast_factor)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return adjust_contrastv2_eager_fallback(
images, contrast_factor, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"AdjustContrastv2", images=images, contrast_factor=contrast_factor,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"AdjustContrastv2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
AdjustContrastv2 = tf_export("raw_ops.AdjustContrastv2")(_ops.to_raw_op(adjust_contrastv2))
def adjust_contrastv2_eager_fallback(images: Annotated[Any, TV_AdjustContrastv2_T], contrast_factor: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, TV_AdjustContrastv2_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
contrast_factor = _ops.convert_to_tensor(contrast_factor, _dtypes.float32)
_inputs_flat = [images, contrast_factor]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"AdjustContrastv2", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"AdjustContrastv2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_AdjustHue_T = TypeVar("TV_AdjustHue_T", _atypes.Float32, _atypes.Half)
def adjust_hue(images: Annotated[Any, TV_AdjustHue_T], delta: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, TV_AdjustHue_T]:
r"""Adjust the hue of one or more images.
`images` is a tensor of at least 3 dimensions. The last dimension is
interpreted as channels, and must be three.
The input image is considered in the RGB colorspace. Conceptually, the RGB
colors are first mapped into HSV. A delta is then applied all the hue values,
and then remapped back to RGB colorspace.
Args:
images: A `Tensor`. Must be one of the following types: `half`, `float32`.
Images to adjust. At least 3-D.
delta: A `Tensor` of type `float32`. A float delta to add to the hue.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "AdjustHue", name, images, delta)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return adjust_hue_eager_fallback(
images, delta, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"AdjustHue", images=images, delta=delta, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"AdjustHue", _inputs_flat, _attrs, _result)
_result, = _result
return _result
AdjustHue = tf_export("raw_ops.AdjustHue")(_ops.to_raw_op(adjust_hue))
def adjust_hue_eager_fallback(images: Annotated[Any, TV_AdjustHue_T], delta: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, TV_AdjustHue_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
delta = _ops.convert_to_tensor(delta, _dtypes.float32)
_inputs_flat = [images, delta]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"AdjustHue", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"AdjustHue", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_AdjustSaturation_T = TypeVar("TV_AdjustSaturation_T", _atypes.Float32, _atypes.Half)
def adjust_saturation(images: Annotated[Any, TV_AdjustSaturation_T], scale: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, TV_AdjustSaturation_T]:
r"""Adjust the saturation of one or more images.
`images` is a tensor of at least 3 dimensions. The last dimension is
interpreted as channels, and must be three.
The input image is considered in the RGB colorspace. Conceptually, the RGB
colors are first mapped into HSV. A scale is then applied all the saturation
values, and then remapped back to RGB colorspace.
Args:
images: A `Tensor`. Must be one of the following types: `half`, `float32`.
Images to adjust. At least 3-D.
scale: A `Tensor` of type `float32`.
A float scale to add to the saturation.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "AdjustSaturation", name, images, scale)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return adjust_saturation_eager_fallback(
images, scale, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"AdjustSaturation", images=images, scale=scale, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"AdjustSaturation", _inputs_flat, _attrs, _result)
_result, = _result
return _result
AdjustSaturation = tf_export("raw_ops.AdjustSaturation")(_ops.to_raw_op(adjust_saturation))
def adjust_saturation_eager_fallback(images: Annotated[Any, TV_AdjustSaturation_T], scale: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, TV_AdjustSaturation_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
scale = _ops.convert_to_tensor(scale, _dtypes.float32)
_inputs_flat = [images, scale]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"AdjustSaturation", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"AdjustSaturation", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_CombinedNonMaxSuppressionOutput = collections.namedtuple(
"CombinedNonMaxSuppression",
["nmsed_boxes", "nmsed_scores", "nmsed_classes", "valid_detections"])
def combined_non_max_suppression(boxes: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size_per_class: Annotated[Any, _atypes.Int32], max_total_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, _atypes.Float32], score_threshold: Annotated[Any, _atypes.Float32], pad_per_class:bool=False, clip_boxes:bool=True, name=None):
r"""Greedily selects a subset of bounding boxes in descending order of score,
This operation performs non_max_suppression on the inputs per batch, across
all classes.
Prunes away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system. Also note that
this algorithm is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is the final boxes, scores and classes tensor
returned after performing non_max_suppression.
Args:
boxes: A `Tensor` of type `float32`.
A 4-D float tensor of shape `[batch_size, num_boxes, q, 4]`. If `q` is 1 then
same boxes are used for all classes otherwise, if `q` is equal to number of
classes, class-specific boxes are used.
scores: A `Tensor` of type `float32`.
A 3-D float tensor of shape `[batch_size, num_boxes, num_classes]`
representing a single score corresponding to each box (each row of boxes).
max_output_size_per_class: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression per class
max_total_size: A `Tensor` of type `int32`.
An int32 scalar representing the maximum number of boxes retained over all
classes. Note that setting this value to a large number may result in OOM error
depending on the system workload.
iou_threshold: A `Tensor` of type `float32`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too much with respect to IOU.
score_threshold: A `Tensor` of type `float32`.
A 0-D float tensor representing the threshold for deciding when to remove
boxes based on score.
pad_per_class: An optional `bool`. Defaults to `False`.
If false, the output nmsed boxes, scores and classes
are padded/clipped to `max_total_size`. If true, the
output nmsed boxes, scores and classes are padded to be of length
`max_size_per_class`*`num_classes`, unless it exceeds `max_total_size` in
which case it is clipped to `max_total_size`. Defaults to false.
clip_boxes: An optional `bool`. Defaults to `True`.
If true, assume the box coordinates are between [0, 1] and clip the output boxes
if they fall beyond [0, 1]. If false, do not do clipping and output the box
coordinates as it is.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (nmsed_boxes, nmsed_scores, nmsed_classes, valid_detections).
nmsed_boxes: A `Tensor` of type `float32`.
nmsed_scores: A `Tensor` of type `float32`.
nmsed_classes: A `Tensor` of type `float32`.
valid_detections: A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "CombinedNonMaxSuppression", name, boxes, scores,
max_output_size_per_class, max_total_size, iou_threshold,
score_threshold, "pad_per_class", pad_per_class, "clip_boxes",
clip_boxes)
_result = _CombinedNonMaxSuppressionOutput._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return combined_non_max_suppression_eager_fallback(
boxes, scores, max_output_size_per_class, max_total_size,
iou_threshold, score_threshold, pad_per_class=pad_per_class,
clip_boxes=clip_boxes, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if pad_per_class is None:
pad_per_class = False
pad_per_class = _execute.make_bool(pad_per_class, "pad_per_class")
if clip_boxes is None:
clip_boxes = True
clip_boxes = _execute.make_bool(clip_boxes, "clip_boxes")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"CombinedNonMaxSuppression", boxes=boxes, scores=scores,
max_output_size_per_class=max_output_size_per_class,
max_total_size=max_total_size,
iou_threshold=iou_threshold,
score_threshold=score_threshold,
pad_per_class=pad_per_class,
clip_boxes=clip_boxes, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("pad_per_class", _op._get_attr_bool("pad_per_class"),
"clip_boxes", _op._get_attr_bool("clip_boxes"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"CombinedNonMaxSuppression", _inputs_flat, _attrs, _result)
_result = _CombinedNonMaxSuppressionOutput._make(_result)
return _result
CombinedNonMaxSuppression = tf_export("raw_ops.CombinedNonMaxSuppression")(_ops.to_raw_op(combined_non_max_suppression))
def combined_non_max_suppression_eager_fallback(boxes: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size_per_class: Annotated[Any, _atypes.Int32], max_total_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, _atypes.Float32], score_threshold: Annotated[Any, _atypes.Float32], pad_per_class: bool, clip_boxes: bool, name, ctx):
if pad_per_class is None:
pad_per_class = False
pad_per_class = _execute.make_bool(pad_per_class, "pad_per_class")
if clip_boxes is None:
clip_boxes = True
clip_boxes = _execute.make_bool(clip_boxes, "clip_boxes")
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
scores = _ops.convert_to_tensor(scores, _dtypes.float32)
max_output_size_per_class = _ops.convert_to_tensor(max_output_size_per_class, _dtypes.int32)
max_total_size = _ops.convert_to_tensor(max_total_size, _dtypes.int32)
iou_threshold = _ops.convert_to_tensor(iou_threshold, _dtypes.float32)
score_threshold = _ops.convert_to_tensor(score_threshold, _dtypes.float32)
_inputs_flat = [boxes, scores, max_output_size_per_class, max_total_size, iou_threshold, score_threshold]
_attrs = ("pad_per_class", pad_per_class, "clip_boxes", clip_boxes)
_result = _execute.execute(b"CombinedNonMaxSuppression", 4,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"CombinedNonMaxSuppression", _inputs_flat, _attrs, _result)
_result = _CombinedNonMaxSuppressionOutput._make(_result)
return _result
TV_CropAndResize_T = TypeVar("TV_CropAndResize_T", _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def crop_and_resize(image: Annotated[Any, TV_CropAndResize_T], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], crop_size: Annotated[Any, _atypes.Int32], method:str="bilinear", extrapolation_value:float=0, name=None) -> Annotated[Any, _atypes.Float32]:
r"""Extracts crops from the input image tensor and resizes them.
Extracts crops from the input image tensor and resizes them using bilinear
sampling or nearest neighbor sampling (possibly with aspect ratio change) to a
common output size specified by `crop_size`. This is more general than the
`crop_to_bounding_box` op which extracts a fixed size slice from the input image
and does not allow resizing or aspect ratio change.
Returns a tensor with `crops` from the input `image` at positions defined at the
bounding box locations in `boxes`. The cropped boxes are all resized (with
bilinear or nearest neighbor interpolation) to a fixed
`size = [crop_height, crop_width]`. The result is a 4-D tensor
`[num_boxes, crop_height, crop_width, depth]`. The resizing is corner aligned.
In particular, if `boxes = [[0, 0, 1, 1]]`, the method will give identical
results to using `tf.image.resize_bilinear()` or
`tf.image.resize_nearest_neighbor()`(depends on the `method` argument) with
`align_corners=True`.
Args:
image: A `Tensor`. Must be one of the following types: `uint8`, `uint16`, `int8`, `int16`, `int32`, `int64`, `half`, `float32`, `float64`.
A 4-D tensor of shape `[batch, image_height, image_width, depth]`.
Both `image_height` and `image_width` need to be positive.
boxes: A `Tensor` of type `float32`.
A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified
in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of
`y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the
`[0, 1]` interval of normalized image height is mapped to
`[0, image_height - 1]` in image height coordinates. We do allow `y1` > `y2`, in
which case the sampled crop is an up-down flipped version of the original
image. The width dimension is treated similarly. Normalized coordinates
outside the `[0, 1]` range are allowed, in which case we use
`extrapolation_value` to extrapolate the input image values.
box_ind: A `Tensor` of type `int32`.
A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
crop_size: A `Tensor` of type `int32`.
A 1-D tensor of 2 elements, `size = [crop_height, crop_width]`. All
cropped image patches are resized to this size. The aspect ratio of the image
content is not preserved. Both `crop_height` and `crop_width` need to be
positive.
method: An optional `string` from: `"bilinear", "nearest"`. Defaults to `"bilinear"`.
A string specifying the sampling method for resizing. It can be either
`"bilinear"` or `"nearest"` and default to `"bilinear"`. Currently two sampling
methods are supported: Bilinear and Nearest Neighbor.
extrapolation_value: An optional `float`. Defaults to `0`.
Value used for extrapolation, when applicable.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "CropAndResize", name, image, boxes, box_ind, crop_size,
"method", method, "extrapolation_value", extrapolation_value)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return crop_and_resize_eager_fallback(
image, boxes, box_ind, crop_size, method=method,
extrapolation_value=extrapolation_value, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
if extrapolation_value is None:
extrapolation_value = 0
extrapolation_value = _execute.make_float(extrapolation_value, "extrapolation_value")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"CropAndResize", image=image, boxes=boxes, box_ind=box_ind,
crop_size=crop_size, method=method,
extrapolation_value=extrapolation_value, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "method", _op.get_attr("method"),
"extrapolation_value", _op.get_attr("extrapolation_value"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"CropAndResize", _inputs_flat, _attrs, _result)
_result, = _result
return _result
CropAndResize = tf_export("raw_ops.CropAndResize")(_ops.to_raw_op(crop_and_resize))
def crop_and_resize_eager_fallback(image: Annotated[Any, TV_CropAndResize_T], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], crop_size: Annotated[Any, _atypes.Int32], method: str, extrapolation_value: float, name, ctx) -> Annotated[Any, _atypes.Float32]:
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
if extrapolation_value is None:
extrapolation_value = 0
extrapolation_value = _execute.make_float(extrapolation_value, "extrapolation_value")
_attr_T, (image,) = _execute.args_to_matching_eager([image], ctx, [_dtypes.uint8, _dtypes.uint16, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.float32, _dtypes.float64, ])
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
box_ind = _ops.convert_to_tensor(box_ind, _dtypes.int32)
crop_size = _ops.convert_to_tensor(crop_size, _dtypes.int32)
_inputs_flat = [image, boxes, box_ind, crop_size]
_attrs = ("T", _attr_T, "method", method, "extrapolation_value",
extrapolation_value)
_result = _execute.execute(b"CropAndResize", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"CropAndResize", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_CropAndResizeGradBoxes_T = TypeVar("TV_CropAndResizeGradBoxes_T", _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def crop_and_resize_grad_boxes(grads: Annotated[Any, _atypes.Float32], image: Annotated[Any, TV_CropAndResizeGradBoxes_T], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], method:str="bilinear", name=None) -> Annotated[Any, _atypes.Float32]:
r"""Computes the gradient of the crop_and_resize op wrt the input boxes tensor.
Args:
grads: A `Tensor` of type `float32`.
A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.
image: A `Tensor`. Must be one of the following types: `uint8`, `uint16`, `int8`, `int16`, `int32`, `int64`, `half`, `float32`, `float64`.
A 4-D tensor of shape `[batch, image_height, image_width, depth]`.
Both `image_height` and `image_width` need to be positive.
boxes: A `Tensor` of type `float32`.
A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified
in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of
`y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the
`[0, 1]` interval of normalized image height is mapped to
`[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in
which case the sampled crop is an up-down flipped version of the original
image. The width dimension is treated similarly. Normalized coordinates
outside the `[0, 1]` range are allowed, in which case we use
`extrapolation_value` to extrapolate the input image values.
box_ind: A `Tensor` of type `int32`.
A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
method: An optional `string` from: `"bilinear"`. Defaults to `"bilinear"`.
A string specifying the interpolation method. Only 'bilinear' is
supported for now.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "CropAndResizeGradBoxes", name, grads, image, boxes, box_ind,
"method", method)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return crop_and_resize_grad_boxes_eager_fallback(
grads, image, boxes, box_ind, method=method, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"CropAndResizeGradBoxes", grads=grads, image=image, boxes=boxes,
box_ind=box_ind, method=method, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "method", _op.get_attr("method"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"CropAndResizeGradBoxes", _inputs_flat, _attrs, _result)
_result, = _result
return _result
CropAndResizeGradBoxes = tf_export("raw_ops.CropAndResizeGradBoxes")(_ops.to_raw_op(crop_and_resize_grad_boxes))
def crop_and_resize_grad_boxes_eager_fallback(grads: Annotated[Any, _atypes.Float32], image: Annotated[Any, TV_CropAndResizeGradBoxes_T], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], method: str, name, ctx) -> Annotated[Any, _atypes.Float32]:
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
_attr_T, (image,) = _execute.args_to_matching_eager([image], ctx, [_dtypes.uint8, _dtypes.uint16, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.float32, _dtypes.float64, ])
grads = _ops.convert_to_tensor(grads, _dtypes.float32)
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
box_ind = _ops.convert_to_tensor(box_ind, _dtypes.int32)
_inputs_flat = [grads, image, boxes, box_ind]
_attrs = ("T", _attr_T, "method", method)
_result = _execute.execute(b"CropAndResizeGradBoxes", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"CropAndResizeGradBoxes", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_CropAndResizeGradImage_T = TypeVar("TV_CropAndResizeGradImage_T", _atypes.Float32, _atypes.Float64, _atypes.Half)
def crop_and_resize_grad_image(grads: Annotated[Any, _atypes.Float32], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], image_size: Annotated[Any, _atypes.Int32], T: TV_CropAndResizeGradImage_T, method:str="bilinear", name=None) -> Annotated[Any, TV_CropAndResizeGradImage_T]:
r"""Computes the gradient of the crop_and_resize op wrt the input image tensor.
Args:
grads: A `Tensor` of type `float32`.
A 4-D tensor of shape `[num_boxes, crop_height, crop_width, depth]`.
boxes: A `Tensor` of type `float32`.
A 2-D tensor of shape `[num_boxes, 4]`. The `i`-th row of the tensor
specifies the coordinates of a box in the `box_ind[i]` image and is specified
in normalized coordinates `[y1, x1, y2, x2]`. A normalized coordinate value of
`y` is mapped to the image coordinate at `y * (image_height - 1)`, so as the
`[0, 1]` interval of normalized image height is mapped to
`[0, image_height - 1] in image height coordinates. We do allow y1 > y2, in
which case the sampled crop is an up-down flipped version of the original
image. The width dimension is treated similarly. Normalized coordinates
outside the `[0, 1]` range are allowed, in which case we use
`extrapolation_value` to extrapolate the input image values.
box_ind: A `Tensor` of type `int32`.
A 1-D tensor of shape `[num_boxes]` with int32 values in `[0, batch)`.
The value of `box_ind[i]` specifies the image that the `i`-th box refers to.
image_size: A `Tensor` of type `int32`.
A 1-D tensor with value `[batch, image_height, image_width, depth]`
containing the original image size. Both `image_height` and `image_width` need
to be positive.
T: A `tf.DType` from: `tf.float32, tf.half, tf.float64`.
method: An optional `string` from: `"bilinear", "nearest"`. Defaults to `"bilinear"`.
A string specifying the interpolation method. Only 'bilinear' is
supported for now.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `T`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "CropAndResizeGradImage", name, grads, boxes, box_ind,
image_size, "T", T, "method", method)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return crop_and_resize_grad_image_eager_fallback(
grads, boxes, box_ind, image_size, T=T, method=method, name=name,
ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
T = _execute.make_type(T, "T")
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"CropAndResizeGradImage", grads=grads, boxes=boxes, box_ind=box_ind,
image_size=image_size, T=T, method=method,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "method", _op.get_attr("method"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"CropAndResizeGradImage", _inputs_flat, _attrs, _result)
_result, = _result
return _result
CropAndResizeGradImage = tf_export("raw_ops.CropAndResizeGradImage")(_ops.to_raw_op(crop_and_resize_grad_image))
def crop_and_resize_grad_image_eager_fallback(grads: Annotated[Any, _atypes.Float32], boxes: Annotated[Any, _atypes.Float32], box_ind: Annotated[Any, _atypes.Int32], image_size: Annotated[Any, _atypes.Int32], T: TV_CropAndResizeGradImage_T, method: str, name, ctx) -> Annotated[Any, TV_CropAndResizeGradImage_T]:
T = _execute.make_type(T, "T")
if method is None:
method = "bilinear"
method = _execute.make_str(method, "method")
grads = _ops.convert_to_tensor(grads, _dtypes.float32)
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
box_ind = _ops.convert_to_tensor(box_ind, _dtypes.int32)
image_size = _ops.convert_to_tensor(image_size, _dtypes.int32)
_inputs_flat = [grads, boxes, box_ind, image_size]
_attrs = ("T", T, "method", method)
_result = _execute.execute(b"CropAndResizeGradImage", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"CropAndResizeGradImage", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def decode_and_crop_jpeg(contents: Annotated[Any, _atypes.String], crop_window: Annotated[Any, _atypes.Int32], channels:int=0, ratio:int=1, fancy_upscaling:bool=True, try_recover_truncated:bool=False, acceptable_fraction:float=1, dct_method:str="", name=None) -> Annotated[Any, _atypes.UInt8]:
r"""Decode and Crop a JPEG-encoded image to a uint8 tensor.
The attr `channels` indicates the desired number of color channels for the
decoded image.
Accepted values are:
* 0: Use the number of channels in the JPEG-encoded image.
* 1: output a grayscale image.
* 3: output an RGB image.
If needed, the JPEG-encoded image is transformed to match the requested number
of color channels.
The attr `ratio` allows downscaling the image by an integer factor during
decoding. Allowed values are: 1, 2, 4, and 8. This is much faster than
downscaling the image later.
It is equivalent to a combination of decode and crop, but much faster by only
decoding partial jpeg image.
Args:
contents: A `Tensor` of type `string`. 0-D. The JPEG-encoded image.
crop_window: A `Tensor` of type `int32`.
1-D. The crop window: [crop_y, crop_x, crop_height, crop_width].
channels: An optional `int`. Defaults to `0`.
Number of color channels for the decoded image.
ratio: An optional `int`. Defaults to `1`. Downscaling ratio.
fancy_upscaling: An optional `bool`. Defaults to `True`.
If true use a slower but nicer upscaling of the
chroma planes (yuv420/422 only).
try_recover_truncated: An optional `bool`. Defaults to `False`.
If true try to recover an image from truncated input.
acceptable_fraction: An optional `float`. Defaults to `1`.
The minimum required fraction of lines before a truncated
input is accepted.
dct_method: An optional `string`. Defaults to `""`.
string specifying a hint about the algorithm used for
decompression. Defaults to "" which maps to a system-specific
default. Currently valid values are ["INTEGER_FAST",
"INTEGER_ACCURATE"]. The hint may be ignored (e.g., the internal
jpeg library changes to a version that does not have that specific
option.)
name: A name for the operation (optional).
Returns:
A `Tensor` of type `uint8`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodeAndCropJpeg", name, contents, crop_window, "channels",
channels, "ratio", ratio, "fancy_upscaling", fancy_upscaling,
"try_recover_truncated", try_recover_truncated, "acceptable_fraction",
acceptable_fraction, "dct_method", dct_method)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_and_crop_jpeg_eager_fallback(
contents, crop_window, channels=channels, ratio=ratio,
fancy_upscaling=fancy_upscaling,
try_recover_truncated=try_recover_truncated,
acceptable_fraction=acceptable_fraction, dct_method=dct_method,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if ratio is None:
ratio = 1
ratio = _execute.make_int(ratio, "ratio")
if fancy_upscaling is None:
fancy_upscaling = True
fancy_upscaling = _execute.make_bool(fancy_upscaling, "fancy_upscaling")
if try_recover_truncated is None:
try_recover_truncated = False
try_recover_truncated = _execute.make_bool(try_recover_truncated, "try_recover_truncated")
if acceptable_fraction is None:
acceptable_fraction = 1
acceptable_fraction = _execute.make_float(acceptable_fraction, "acceptable_fraction")
if dct_method is None:
dct_method = ""
dct_method = _execute.make_str(dct_method, "dct_method")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodeAndCropJpeg", contents=contents, crop_window=crop_window,
channels=channels, ratio=ratio,
fancy_upscaling=fancy_upscaling,
try_recover_truncated=try_recover_truncated,
acceptable_fraction=acceptable_fraction,
dct_method=dct_method, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("channels", _op._get_attr_int("channels"), "ratio",
_op._get_attr_int("ratio"), "fancy_upscaling",
_op._get_attr_bool("fancy_upscaling"), "try_recover_truncated",
_op._get_attr_bool("try_recover_truncated"),
"acceptable_fraction", _op.get_attr("acceptable_fraction"),
"dct_method", _op.get_attr("dct_method"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodeAndCropJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodeAndCropJpeg = tf_export("raw_ops.DecodeAndCropJpeg")(_ops.to_raw_op(decode_and_crop_jpeg))
def decode_and_crop_jpeg_eager_fallback(contents: Annotated[Any, _atypes.String], crop_window: Annotated[Any, _atypes.Int32], channels: int, ratio: int, fancy_upscaling: bool, try_recover_truncated: bool, acceptable_fraction: float, dct_method: str, name, ctx) -> Annotated[Any, _atypes.UInt8]:
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if ratio is None:
ratio = 1
ratio = _execute.make_int(ratio, "ratio")
if fancy_upscaling is None:
fancy_upscaling = True
fancy_upscaling = _execute.make_bool(fancy_upscaling, "fancy_upscaling")
if try_recover_truncated is None:
try_recover_truncated = False
try_recover_truncated = _execute.make_bool(try_recover_truncated, "try_recover_truncated")
if acceptable_fraction is None:
acceptable_fraction = 1
acceptable_fraction = _execute.make_float(acceptable_fraction, "acceptable_fraction")
if dct_method is None:
dct_method = ""
dct_method = _execute.make_str(dct_method, "dct_method")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
crop_window = _ops.convert_to_tensor(crop_window, _dtypes.int32)
_inputs_flat = [contents, crop_window]
_attrs = ("channels", channels, "ratio", ratio, "fancy_upscaling",
fancy_upscaling, "try_recover_truncated", try_recover_truncated,
"acceptable_fraction", acceptable_fraction, "dct_method", dct_method)
_result = _execute.execute(b"DecodeAndCropJpeg", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodeAndCropJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def decode_bmp(contents: Annotated[Any, _atypes.String], channels:int=0, name=None) -> Annotated[Any, _atypes.UInt8]:
r"""Decode the first frame of a BMP-encoded image to a uint8 tensor.
The attr `channels` indicates the desired number of color channels for the
decoded image.
Accepted values are:
* 0: Use the number of channels in the BMP-encoded image.
* 3: output an RGB image.
* 4: output an RGBA image.
Args:
contents: A `Tensor` of type `string`. 0-D. The BMP-encoded image.
channels: An optional `int`. Defaults to `0`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `uint8`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodeBmp", name, contents, "channels", channels)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_bmp_eager_fallback(
contents, channels=channels, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodeBmp", contents=contents, channels=channels, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("channels", _op._get_attr_int("channels"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodeBmp", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodeBmp = tf_export("raw_ops.DecodeBmp")(_ops.to_raw_op(decode_bmp))
def decode_bmp_eager_fallback(contents: Annotated[Any, _atypes.String], channels: int, name, ctx) -> Annotated[Any, _atypes.UInt8]:
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = ("channels", channels)
_result = _execute.execute(b"DecodeBmp", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodeBmp", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def decode_gif(contents: Annotated[Any, _atypes.String], name=None) -> Annotated[Any, _atypes.UInt8]:
r"""Decode the frame(s) of a GIF-encoded image to a uint8 tensor.
GIF images with frame or transparency compression are not supported.
On Linux and MacOS systems, convert animated GIFs from compressed to
uncompressed by running:
convert $src.gif -coalesce $dst.gif
This op also supports decoding JPEGs and PNGs, though it is cleaner to use
`tf.io.decode_image`.
Args:
contents: A `Tensor` of type `string`. 0-D. The GIF-encoded image.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `uint8`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodeGif", name, contents)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_gif_eager_fallback(
contents, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodeGif", contents=contents, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ()
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodeGif", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodeGif = tf_export("raw_ops.DecodeGif")(_ops.to_raw_op(decode_gif))
def decode_gif_eager_fallback(contents: Annotated[Any, _atypes.String], name, ctx) -> Annotated[Any, _atypes.UInt8]:
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = None
_result = _execute.execute(b"DecodeGif", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodeGif", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_DecodeImage_dtype = TypeVar("TV_DecodeImage_dtype", _atypes.Float32, _atypes.UInt16, _atypes.UInt8)
def decode_image(contents: Annotated[Any, _atypes.String], channels:int=0, dtype:TV_DecodeImage_dtype=_dtypes.uint8, expand_animations:bool=True, name=None) -> Annotated[Any, TV_DecodeImage_dtype]:
r"""Function for decode_bmp, decode_gif, decode_jpeg, and decode_png.
Detects whether an image is a BMP, GIF, JPEG, or PNG, and performs the
appropriate operation to convert the input bytes string into a Tensor of type
dtype.
*NOTE*: decode_gif returns a 4-D array [num_frames, height, width, 3], as
opposed to decode_bmp, decode_jpeg and decode_png, which return 3-D arrays
[height, width, num_channels]. Make sure to take this into account when
constructing your graph if you are intermixing GIF files with BMP, JPEG, and/or
PNG files. Alternately, set the expand_animations argument of this function to
False, in which case the op will return 3-dimensional tensors and will truncate
animated GIF files to the first frame.
*NOTE*: If the first frame of an animated GIF does not occupy the entire
canvas (maximum frame width x maximum frame height), then it fills the
unoccupied areas (in the first frame) with zeros (black). For frames after the
first frame that does not occupy the entire canvas, it uses the previous
frame to fill the unoccupied areas.
Args:
contents: A `Tensor` of type `string`. 0-D. The encoded image bytes.
channels: An optional `int`. Defaults to `0`.
Number of color channels for the decoded image.
dtype: An optional `tf.DType` from: `tf.uint8, tf.uint16, tf.float32`. Defaults to `tf.uint8`.
The desired DType of the returned Tensor.
expand_animations: An optional `bool`. Defaults to `True`.
Controls the output shape of the returned op. If True, the returned op will
produce a 3-D tensor for PNG, JPEG, and BMP files; and a 4-D tensor for all
GIFs, whether animated or not. If, False, the returned op will produce a 3-D
tensor for all file types and will truncate animated GIFs to the first frame.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `dtype`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodeImage", name, contents, "channels", channels, "dtype",
dtype, "expand_animations", expand_animations)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_image_eager_fallback(
contents, channels=channels, dtype=dtype,
expand_animations=expand_animations, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if dtype is None:
dtype = _dtypes.uint8
dtype = _execute.make_type(dtype, "dtype")
if expand_animations is None:
expand_animations = True
expand_animations = _execute.make_bool(expand_animations, "expand_animations")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodeImage", contents=contents, channels=channels, dtype=dtype,
expand_animations=expand_animations, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("channels", _op._get_attr_int("channels"), "dtype",
_op._get_attr_type("dtype"), "expand_animations",
_op._get_attr_bool("expand_animations"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodeImage", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodeImage = tf_export("raw_ops.DecodeImage")(_ops.to_raw_op(decode_image))
def decode_image_eager_fallback(contents: Annotated[Any, _atypes.String], channels: int, dtype: TV_DecodeImage_dtype, expand_animations: bool, name, ctx) -> Annotated[Any, TV_DecodeImage_dtype]:
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if dtype is None:
dtype = _dtypes.uint8
dtype = _execute.make_type(dtype, "dtype")
if expand_animations is None:
expand_animations = True
expand_animations = _execute.make_bool(expand_animations, "expand_animations")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = ("channels", channels, "dtype", dtype, "expand_animations",
expand_animations)
_result = _execute.execute(b"DecodeImage", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodeImage", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def decode_jpeg(contents: Annotated[Any, _atypes.String], channels:int=0, ratio:int=1, fancy_upscaling:bool=True, try_recover_truncated:bool=False, acceptable_fraction:float=1, dct_method:str="", name=None) -> Annotated[Any, _atypes.UInt8]:
r"""Decode a JPEG-encoded image to a uint8 tensor.
The attr `channels` indicates the desired number of color channels for the
decoded image.
Accepted values are:
* 0: Use the number of channels in the JPEG-encoded image.
* 1: output a grayscale image.
* 3: output an RGB image.
If needed, the JPEG-encoded image is transformed to match the requested number
of color channels.
The attr `ratio` allows downscaling the image by an integer factor during
decoding. Allowed values are: 1, 2, 4, and 8. This is much faster than
downscaling the image later.
This op also supports decoding PNGs and non-animated GIFs since the interface is
the same, though it is cleaner to use `tf.io.decode_image`.
Args:
contents: A `Tensor` of type `string`. 0-D. The JPEG-encoded image.
channels: An optional `int`. Defaults to `0`.
Number of color channels for the decoded image.
ratio: An optional `int`. Defaults to `1`. Downscaling ratio.
fancy_upscaling: An optional `bool`. Defaults to `True`.
If true use a slower but nicer upscaling of the
chroma planes (yuv420/422 only).
try_recover_truncated: An optional `bool`. Defaults to `False`.
If true try to recover an image from truncated input.
acceptable_fraction: An optional `float`. Defaults to `1`.
The minimum required fraction of lines before a truncated
input is accepted.
dct_method: An optional `string`. Defaults to `""`.
string specifying a hint about the algorithm used for
decompression. Defaults to "" which maps to a system-specific
default. Currently valid values are ["INTEGER_FAST",
"INTEGER_ACCURATE"]. The hint may be ignored (e.g., the internal
jpeg library changes to a version that does not have that specific
option.)
name: A name for the operation (optional).
Returns:
A `Tensor` of type `uint8`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodeJpeg", name, contents, "channels", channels, "ratio",
ratio, "fancy_upscaling", fancy_upscaling, "try_recover_truncated",
try_recover_truncated, "acceptable_fraction", acceptable_fraction,
"dct_method", dct_method)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_jpeg_eager_fallback(
contents, channels=channels, ratio=ratio,
fancy_upscaling=fancy_upscaling,
try_recover_truncated=try_recover_truncated,
acceptable_fraction=acceptable_fraction, dct_method=dct_method,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if ratio is None:
ratio = 1
ratio = _execute.make_int(ratio, "ratio")
if fancy_upscaling is None:
fancy_upscaling = True
fancy_upscaling = _execute.make_bool(fancy_upscaling, "fancy_upscaling")
if try_recover_truncated is None:
try_recover_truncated = False
try_recover_truncated = _execute.make_bool(try_recover_truncated, "try_recover_truncated")
if acceptable_fraction is None:
acceptable_fraction = 1
acceptable_fraction = _execute.make_float(acceptable_fraction, "acceptable_fraction")
if dct_method is None:
dct_method = ""
dct_method = _execute.make_str(dct_method, "dct_method")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodeJpeg", contents=contents, channels=channels, ratio=ratio,
fancy_upscaling=fancy_upscaling,
try_recover_truncated=try_recover_truncated,
acceptable_fraction=acceptable_fraction,
dct_method=dct_method, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("channels", _op._get_attr_int("channels"), "ratio",
_op._get_attr_int("ratio"), "fancy_upscaling",
_op._get_attr_bool("fancy_upscaling"), "try_recover_truncated",
_op._get_attr_bool("try_recover_truncated"),
"acceptable_fraction", _op.get_attr("acceptable_fraction"),
"dct_method", _op.get_attr("dct_method"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodeJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodeJpeg = tf_export("raw_ops.DecodeJpeg")(_ops.to_raw_op(decode_jpeg))
def decode_jpeg_eager_fallback(contents: Annotated[Any, _atypes.String], channels: int, ratio: int, fancy_upscaling: bool, try_recover_truncated: bool, acceptable_fraction: float, dct_method: str, name, ctx) -> Annotated[Any, _atypes.UInt8]:
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if ratio is None:
ratio = 1
ratio = _execute.make_int(ratio, "ratio")
if fancy_upscaling is None:
fancy_upscaling = True
fancy_upscaling = _execute.make_bool(fancy_upscaling, "fancy_upscaling")
if try_recover_truncated is None:
try_recover_truncated = False
try_recover_truncated = _execute.make_bool(try_recover_truncated, "try_recover_truncated")
if acceptable_fraction is None:
acceptable_fraction = 1
acceptable_fraction = _execute.make_float(acceptable_fraction, "acceptable_fraction")
if dct_method is None:
dct_method = ""
dct_method = _execute.make_str(dct_method, "dct_method")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = ("channels", channels, "ratio", ratio, "fancy_upscaling",
fancy_upscaling, "try_recover_truncated", try_recover_truncated,
"acceptable_fraction", acceptable_fraction, "dct_method", dct_method)
_result = _execute.execute(b"DecodeJpeg", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodeJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_DecodePng_dtype = TypeVar("TV_DecodePng_dtype", _atypes.UInt16, _atypes.UInt8)
def decode_png(contents: Annotated[Any, _atypes.String], channels:int=0, dtype:TV_DecodePng_dtype=_dtypes.uint8, name=None) -> Annotated[Any, TV_DecodePng_dtype]:
r"""Decode a PNG-encoded image to a uint8 or uint16 tensor.
The attr `channels` indicates the desired number of color channels for the
decoded image.
Accepted values are:
* 0: Use the number of channels in the PNG-encoded image.
* 1: output a grayscale image.
* 3: output an RGB image.
* 4: output an RGBA image.
If needed, the PNG-encoded image is transformed to match the requested number
of color channels.
This op also supports decoding JPEGs and non-animated GIFs since the interface
is the same, though it is cleaner to use `tf.io.decode_image`.
Args:
contents: A `Tensor` of type `string`. 0-D. The PNG-encoded image.
channels: An optional `int`. Defaults to `0`.
Number of color channels for the decoded image.
dtype: An optional `tf.DType` from: `tf.uint8, tf.uint16`. Defaults to `tf.uint8`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `dtype`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DecodePng", name, contents, "channels", channels, "dtype",
dtype)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return decode_png_eager_fallback(
contents, channels=channels, dtype=dtype, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if dtype is None:
dtype = _dtypes.uint8
dtype = _execute.make_type(dtype, "dtype")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DecodePng", contents=contents, channels=channels, dtype=dtype,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("channels", _op._get_attr_int("channels"), "dtype",
_op._get_attr_type("dtype"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DecodePng", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DecodePng = tf_export("raw_ops.DecodePng")(_ops.to_raw_op(decode_png))
def decode_png_eager_fallback(contents: Annotated[Any, _atypes.String], channels: int, dtype: TV_DecodePng_dtype, name, ctx) -> Annotated[Any, TV_DecodePng_dtype]:
if channels is None:
channels = 0
channels = _execute.make_int(channels, "channels")
if dtype is None:
dtype = _dtypes.uint8
dtype = _execute.make_type(dtype, "dtype")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = ("channels", channels, "dtype", dtype)
_result = _execute.execute(b"DecodePng", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DecodePng", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_DrawBoundingBoxes_T = TypeVar("TV_DrawBoundingBoxes_T", _atypes.Float32, _atypes.Half)
def draw_bounding_boxes(images: Annotated[Any, TV_DrawBoundingBoxes_T], boxes: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, TV_DrawBoundingBoxes_T]:
r"""Draw bounding boxes on a batch of images.
Outputs a copy of `images` but draws on top of the pixels zero or more bounding
boxes specified by the locations in `boxes`. The coordinates of the each
bounding box in `boxes` are encoded as `[y_min, x_min, y_max, x_max]`. The
bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and
height of the underlying image.
For example, if an image is 100 x 200 pixels (height x width) and the bounding
box is `[0.1, 0.2, 0.5, 0.9]`, the upper-left and bottom-right coordinates of
the bounding box will be `(40, 10)` to `(180, 50)` (in (x,y) coordinates).
Parts of the bounding box may fall outside the image.
Args:
images: A `Tensor`. Must be one of the following types: `float32`, `half`.
4-D with shape `[batch, height, width, depth]`. A batch of images.
boxes: A `Tensor` of type `float32`.
3-D with shape `[batch, num_bounding_boxes, 4]` containing bounding
boxes.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DrawBoundingBoxes", name, images, boxes)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return draw_bounding_boxes_eager_fallback(
images, boxes, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DrawBoundingBoxes", images=images, boxes=boxes, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DrawBoundingBoxes", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DrawBoundingBoxes = tf_export("raw_ops.DrawBoundingBoxes")(_ops.to_raw_op(draw_bounding_boxes))
def draw_bounding_boxes_eager_fallback(images: Annotated[Any, TV_DrawBoundingBoxes_T], boxes: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, TV_DrawBoundingBoxes_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.float32, _dtypes.half, ], _dtypes.float32)
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
_inputs_flat = [images, boxes]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"DrawBoundingBoxes", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DrawBoundingBoxes", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_DrawBoundingBoxesV2_T = TypeVar("TV_DrawBoundingBoxesV2_T", _atypes.Float32, _atypes.Half)
def draw_bounding_boxes_v2(images: Annotated[Any, TV_DrawBoundingBoxesV2_T], boxes: Annotated[Any, _atypes.Float32], colors: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, TV_DrawBoundingBoxesV2_T]:
r"""Draw bounding boxes on a batch of images.
Outputs a copy of `images` but draws on top of the pixels zero or more bounding
boxes specified by the locations in `boxes`. The coordinates of the each
bounding box in `boxes` are encoded as `[y_min, x_min, y_max, x_max]`. The
bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and
height of the underlying image.
For example, if an image is 100 x 200 pixels (height x width) and the bounding
box is `[0.1, 0.2, 0.5, 0.9]`, the upper-left and bottom-right coordinates of
the bounding box will be `(40, 10)` to `(100, 50)` (in (x,y) coordinates).
Parts of the bounding box may fall outside the image.
Args:
images: A `Tensor`. Must be one of the following types: `float32`, `half`.
4-D with shape `[batch, height, width, depth]`. A batch of images.
boxes: A `Tensor` of type `float32`.
3-D with shape `[batch, num_bounding_boxes, 4]` containing bounding
boxes.
colors: A `Tensor` of type `float32`.
2-D. A list of RGBA colors to cycle through for the boxes.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "DrawBoundingBoxesV2", name, images, boxes, colors)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return draw_bounding_boxes_v2_eager_fallback(
images, boxes, colors, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"DrawBoundingBoxesV2", images=images, boxes=boxes, colors=colors,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"DrawBoundingBoxesV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
DrawBoundingBoxesV2 = tf_export("raw_ops.DrawBoundingBoxesV2")(_ops.to_raw_op(draw_bounding_boxes_v2))
def draw_bounding_boxes_v2_eager_fallback(images: Annotated[Any, TV_DrawBoundingBoxesV2_T], boxes: Annotated[Any, _atypes.Float32], colors: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, TV_DrawBoundingBoxesV2_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.float32, _dtypes.half, ], _dtypes.float32)
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
colors = _ops.convert_to_tensor(colors, _dtypes.float32)
_inputs_flat = [images, boxes, colors]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"DrawBoundingBoxesV2", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"DrawBoundingBoxesV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def encode_jpeg(image: Annotated[Any, _atypes.UInt8], format:str="", quality:int=95, progressive:bool=False, optimize_size:bool=False, chroma_downsampling:bool=True, density_unit:str="in", x_density:int=300, y_density:int=300, xmp_metadata:str="", name=None) -> Annotated[Any, _atypes.String]:
r"""JPEG-encode an image.
`image` is a 3-D uint8 Tensor of shape `[height, width, channels]`.
The attr `format` can be used to override the color format of the encoded
output. Values can be:
* `''`: Use a default format based on the number of channels in the image.
* `grayscale`: Output a grayscale JPEG image. The `channels` dimension
of `image` must be 1.
* `rgb`: Output an RGB JPEG image. The `channels` dimension
of `image` must be 3.
If `format` is not specified or is the empty string, a default format is picked
in function of the number of channels in `image`:
* 1: Output a grayscale image.
* 3: Output an RGB image.
Args:
image: A `Tensor` of type `uint8`.
3-D with shape `[height, width, channels]`.
format: An optional `string` from: `"", "grayscale", "rgb"`. Defaults to `""`.
Per pixel image format.
quality: An optional `int`. Defaults to `95`.
Quality of the compression from 0 to 100 (higher is better and slower).
progressive: An optional `bool`. Defaults to `False`.
If True, create a JPEG that loads progressively (coarse to fine).
optimize_size: An optional `bool`. Defaults to `False`.
If True, spend CPU/RAM to reduce size with no quality change.
chroma_downsampling: An optional `bool`. Defaults to `True`.
See http://en.wikipedia.org/wiki/Chroma_subsampling.
density_unit: An optional `string` from: `"in", "cm"`. Defaults to `"in"`.
Unit used to specify `x_density` and `y_density`:
pixels per inch (`'in'`) or centimeter (`'cm'`).
x_density: An optional `int`. Defaults to `300`.
Horizontal pixels per density unit.
y_density: An optional `int`. Defaults to `300`.
Vertical pixels per density unit.
xmp_metadata: An optional `string`. Defaults to `""`.
If not empty, embed this XMP metadata in the image header.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `string`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "EncodeJpeg", name, image, "format", format, "quality", quality,
"progressive", progressive, "optimize_size", optimize_size,
"chroma_downsampling", chroma_downsampling, "density_unit",
density_unit, "x_density", x_density, "y_density", y_density,
"xmp_metadata", xmp_metadata)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return encode_jpeg_eager_fallback(
image, format=format, quality=quality, progressive=progressive,
optimize_size=optimize_size,
chroma_downsampling=chroma_downsampling, density_unit=density_unit,
x_density=x_density, y_density=y_density, xmp_metadata=xmp_metadata,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if format is None:
format = ""
format = _execute.make_str(format, "format")
if quality is None:
quality = 95
quality = _execute.make_int(quality, "quality")
if progressive is None:
progressive = False
progressive = _execute.make_bool(progressive, "progressive")
if optimize_size is None:
optimize_size = False
optimize_size = _execute.make_bool(optimize_size, "optimize_size")
if chroma_downsampling is None:
chroma_downsampling = True
chroma_downsampling = _execute.make_bool(chroma_downsampling, "chroma_downsampling")
if density_unit is None:
density_unit = "in"
density_unit = _execute.make_str(density_unit, "density_unit")
if x_density is None:
x_density = 300
x_density = _execute.make_int(x_density, "x_density")
if y_density is None:
y_density = 300
y_density = _execute.make_int(y_density, "y_density")
if xmp_metadata is None:
xmp_metadata = ""
xmp_metadata = _execute.make_str(xmp_metadata, "xmp_metadata")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"EncodeJpeg", image=image, format=format, quality=quality,
progressive=progressive, optimize_size=optimize_size,
chroma_downsampling=chroma_downsampling,
density_unit=density_unit, x_density=x_density,
y_density=y_density, xmp_metadata=xmp_metadata,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("format", _op.get_attr("format"), "quality",
_op._get_attr_int("quality"), "progressive",
_op._get_attr_bool("progressive"), "optimize_size",
_op._get_attr_bool("optimize_size"), "chroma_downsampling",
_op._get_attr_bool("chroma_downsampling"), "density_unit",
_op.get_attr("density_unit"), "x_density",
_op._get_attr_int("x_density"), "y_density",
_op._get_attr_int("y_density"), "xmp_metadata",
_op.get_attr("xmp_metadata"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"EncodeJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
EncodeJpeg = tf_export("raw_ops.EncodeJpeg")(_ops.to_raw_op(encode_jpeg))
def encode_jpeg_eager_fallback(image: Annotated[Any, _atypes.UInt8], format: str, quality: int, progressive: bool, optimize_size: bool, chroma_downsampling: bool, density_unit: str, x_density: int, y_density: int, xmp_metadata: str, name, ctx) -> Annotated[Any, _atypes.String]:
if format is None:
format = ""
format = _execute.make_str(format, "format")
if quality is None:
quality = 95
quality = _execute.make_int(quality, "quality")
if progressive is None:
progressive = False
progressive = _execute.make_bool(progressive, "progressive")
if optimize_size is None:
optimize_size = False
optimize_size = _execute.make_bool(optimize_size, "optimize_size")
if chroma_downsampling is None:
chroma_downsampling = True
chroma_downsampling = _execute.make_bool(chroma_downsampling, "chroma_downsampling")
if density_unit is None:
density_unit = "in"
density_unit = _execute.make_str(density_unit, "density_unit")
if x_density is None:
x_density = 300
x_density = _execute.make_int(x_density, "x_density")
if y_density is None:
y_density = 300
y_density = _execute.make_int(y_density, "y_density")
if xmp_metadata is None:
xmp_metadata = ""
xmp_metadata = _execute.make_str(xmp_metadata, "xmp_metadata")
image = _ops.convert_to_tensor(image, _dtypes.uint8)
_inputs_flat = [image]
_attrs = ("format", format, "quality", quality, "progressive", progressive,
"optimize_size", optimize_size, "chroma_downsampling", chroma_downsampling,
"density_unit", density_unit, "x_density", x_density, "y_density",
y_density, "xmp_metadata", xmp_metadata)
_result = _execute.execute(b"EncodeJpeg", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"EncodeJpeg", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def encode_jpeg_variable_quality(images: Annotated[Any, _atypes.UInt8], quality: Annotated[Any, _atypes.Int32], name=None) -> Annotated[Any, _atypes.String]:
r"""JPEG encode input image with provided compression quality.
`image` is a 3-D uint8 Tensor of shape `[height, width, channels]`.
`quality` is an int32 jpeg compression quality value between 0 and 100.
Args:
images: A `Tensor` of type `uint8`. Images to adjust. At least 3-D.
quality: A `Tensor` of type `int32`. An int quality to encode to.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `string`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "EncodeJpegVariableQuality", name, images, quality)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return encode_jpeg_variable_quality_eager_fallback(
images, quality, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"EncodeJpegVariableQuality", images=images, quality=quality,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ()
_inputs_flat = _op.inputs
_execute.record_gradient(
"EncodeJpegVariableQuality", _inputs_flat, _attrs, _result)
_result, = _result
return _result
EncodeJpegVariableQuality = tf_export("raw_ops.EncodeJpegVariableQuality")(_ops.to_raw_op(encode_jpeg_variable_quality))
def encode_jpeg_variable_quality_eager_fallback(images: Annotated[Any, _atypes.UInt8], quality: Annotated[Any, _atypes.Int32], name, ctx) -> Annotated[Any, _atypes.String]:
images = _ops.convert_to_tensor(images, _dtypes.uint8)
quality = _ops.convert_to_tensor(quality, _dtypes.int32)
_inputs_flat = [images, quality]
_attrs = None
_result = _execute.execute(b"EncodeJpegVariableQuality", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"EncodeJpegVariableQuality", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_EncodePng_T = TypeVar("TV_EncodePng_T", _atypes.UInt16, _atypes.UInt8)
def encode_png(image: Annotated[Any, TV_EncodePng_T], compression:int=-1, name=None) -> Annotated[Any, _atypes.String]:
r"""PNG-encode an image.
`image` is a 3-D uint8 or uint16 Tensor of shape `[height, width, channels]`
where `channels` is:
* 1: for grayscale.
* 2: for grayscale + alpha.
* 3: for RGB.
* 4: for RGBA.
The ZLIB compression level, `compression`, can be -1 for the PNG-encoder
default or a value from 0 to 9. 9 is the highest compression level, generating
the smallest output, but is slower.
Args:
image: A `Tensor`. Must be one of the following types: `uint8`, `uint16`.
3-D with shape `[height, width, channels]`.
compression: An optional `int`. Defaults to `-1`. Compression level.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `string`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "EncodePng", name, image, "compression", compression)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return encode_png_eager_fallback(
image, compression=compression, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if compression is None:
compression = -1
compression = _execute.make_int(compression, "compression")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"EncodePng", image=image, compression=compression, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("compression", _op._get_attr_int("compression"), "T",
_op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"EncodePng", _inputs_flat, _attrs, _result)
_result, = _result
return _result
EncodePng = tf_export("raw_ops.EncodePng")(_ops.to_raw_op(encode_png))
def encode_png_eager_fallback(image: Annotated[Any, TV_EncodePng_T], compression: int, name, ctx) -> Annotated[Any, _atypes.String]:
if compression is None:
compression = -1
compression = _execute.make_int(compression, "compression")
_attr_T, (image,) = _execute.args_to_matching_eager([image], ctx, [_dtypes.uint8, _dtypes.uint16, ], _dtypes.uint8)
_inputs_flat = [image]
_attrs = ("compression", compression, "T", _attr_T)
_result = _execute.execute(b"EncodePng", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"EncodePng", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def extract_glimpse(input: Annotated[Any, _atypes.Float32], size: Annotated[Any, _atypes.Int32], offsets: Annotated[Any, _atypes.Float32], centered:bool=True, normalized:bool=True, uniform_noise:bool=True, noise:str="uniform", name=None) -> Annotated[Any, _atypes.Float32]:
r"""Extracts a glimpse from the input tensor.
Returns a set of windows called glimpses extracted at location
`offsets` from the input tensor. If the windows only partially
overlaps the inputs, the non overlapping areas will be filled with
random noise.
The result is a 4-D tensor of shape `[batch_size, glimpse_height,
glimpse_width, channels]`. The channels and batch dimensions are the
same as that of the input tensor. The height and width of the output
windows are specified in the `size` parameter.
The argument `normalized` and `centered` controls how the windows are built:
* If the coordinates are normalized but not centered, 0.0 and 1.0
correspond to the minimum and maximum of each height and width
dimension.
* If the coordinates are both normalized and centered, they range from
-1.0 to 1.0. The coordinates (-1.0, -1.0) correspond to the upper
left corner, the lower right corner is located at (1.0, 1.0) and the
center is at (0, 0).
* If the coordinates are not normalized they are interpreted as
numbers of pixels.
Args:
input: A `Tensor` of type `float32`.
A 4-D float tensor of shape `[batch_size, height, width, channels]`.
size: A `Tensor` of type `int32`.
A 1-D tensor of 2 elements containing the size of the glimpses
to extract. The glimpse height must be specified first, following
by the glimpse width.
offsets: A `Tensor` of type `float32`.
A 2-D integer tensor of shape `[batch_size, 2]` containing
the y, x locations of the center of each window.
centered: An optional `bool`. Defaults to `True`.
indicates if the offset coordinates are centered relative to
the image, in which case the (0, 0) offset is relative to the center
of the input images. If false, the (0,0) offset corresponds to the
upper left corner of the input images.
normalized: An optional `bool`. Defaults to `True`.
indicates if the offset coordinates are normalized.
uniform_noise: An optional `bool`. Defaults to `True`.
indicates if the noise should be generated using a
uniform distribution or a Gaussian distribution.
noise: An optional `string`. Defaults to `"uniform"`.
indicates if the noise should `uniform`, `gaussian`, or
`zero`. The default is `uniform` which means the noise type
will be decided by `uniform_noise`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ExtractGlimpse", name, input, size, offsets, "centered",
centered, "normalized", normalized, "uniform_noise", uniform_noise,
"noise", noise)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return extract_glimpse_eager_fallback(
input, size, offsets, centered=centered, normalized=normalized,
uniform_noise=uniform_noise, noise=noise, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if centered is None:
centered = True
centered = _execute.make_bool(centered, "centered")
if normalized is None:
normalized = True
normalized = _execute.make_bool(normalized, "normalized")
if uniform_noise is None:
uniform_noise = True
uniform_noise = _execute.make_bool(uniform_noise, "uniform_noise")
if noise is None:
noise = "uniform"
noise = _execute.make_str(noise, "noise")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ExtractGlimpse", input=input, size=size, offsets=offsets,
centered=centered, normalized=normalized,
uniform_noise=uniform_noise, noise=noise, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("centered", _op._get_attr_bool("centered"), "normalized",
_op._get_attr_bool("normalized"), "uniform_noise",
_op._get_attr_bool("uniform_noise"), "noise",
_op.get_attr("noise"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ExtractGlimpse", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ExtractGlimpse = tf_export("raw_ops.ExtractGlimpse")(_ops.to_raw_op(extract_glimpse))
def extract_glimpse_eager_fallback(input: Annotated[Any, _atypes.Float32], size: Annotated[Any, _atypes.Int32], offsets: Annotated[Any, _atypes.Float32], centered: bool, normalized: bool, uniform_noise: bool, noise: str, name, ctx) -> Annotated[Any, _atypes.Float32]:
if centered is None:
centered = True
centered = _execute.make_bool(centered, "centered")
if normalized is None:
normalized = True
normalized = _execute.make_bool(normalized, "normalized")
if uniform_noise is None:
uniform_noise = True
uniform_noise = _execute.make_bool(uniform_noise, "uniform_noise")
if noise is None:
noise = "uniform"
noise = _execute.make_str(noise, "noise")
input = _ops.convert_to_tensor(input, _dtypes.float32)
size = _ops.convert_to_tensor(size, _dtypes.int32)
offsets = _ops.convert_to_tensor(offsets, _dtypes.float32)
_inputs_flat = [input, size, offsets]
_attrs = ("centered", centered, "normalized", normalized, "uniform_noise",
uniform_noise, "noise", noise)
_result = _execute.execute(b"ExtractGlimpse", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ExtractGlimpse", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def extract_glimpse_v2(input: Annotated[Any, _atypes.Float32], size: Annotated[Any, _atypes.Int32], offsets: Annotated[Any, _atypes.Float32], centered:bool=True, normalized:bool=True, uniform_noise:bool=True, noise:str="uniform", name=None) -> Annotated[Any, _atypes.Float32]:
r"""Extracts a glimpse from the input tensor.
Returns a set of windows called glimpses extracted at location
`offsets` from the input tensor. If the windows only partially
overlaps the inputs, the non overlapping areas will be filled with
random noise.
The result is a 4-D tensor of shape `[batch_size, glimpse_height,
glimpse_width, channels]`. The channels and batch dimensions are the
same as that of the input tensor. The height and width of the output
windows are specified in the `size` parameter.
The argument `normalized` and `centered` controls how the windows are built:
* If the coordinates are normalized but not centered, 0.0 and 1.0
correspond to the minimum and maximum of each height and width
dimension.
* If the coordinates are both normalized and centered, they range from
-1.0 to 1.0. The coordinates (-1.0, -1.0) correspond to the upper
left corner, the lower right corner is located at (1.0, 1.0) and the
center is at (0, 0).
* If the coordinates are not normalized they are interpreted as
numbers of pixels.
Args:
input: A `Tensor` of type `float32`.
A 4-D float tensor of shape `[batch_size, height, width, channels]`.
size: A `Tensor` of type `int32`.
A 1-D tensor of 2 elements containing the size of the glimpses
to extract. The glimpse height must be specified first, following
by the glimpse width.
offsets: A `Tensor` of type `float32`.
A 2-D integer tensor of shape `[batch_size, 2]` containing
the y, x locations of the center of each window.
centered: An optional `bool`. Defaults to `True`.
indicates if the offset coordinates are centered relative to
the image, in which case the (0, 0) offset is relative to the center
of the input images. If false, the (0,0) offset corresponds to the
upper left corner of the input images.
normalized: An optional `bool`. Defaults to `True`.
indicates if the offset coordinates are normalized.
uniform_noise: An optional `bool`. Defaults to `True`.
indicates if the noise should be generated using a
uniform distribution or a Gaussian distribution.
noise: An optional `string`. Defaults to `"uniform"`.
indicates if the noise should `uniform`, `gaussian`, or
`zero`. The default is `uniform` which means the noise type
will be decided by `uniform_noise`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ExtractGlimpseV2", name, input, size, offsets, "centered",
centered, "normalized", normalized, "uniform_noise", uniform_noise,
"noise", noise)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return extract_glimpse_v2_eager_fallback(
input, size, offsets, centered=centered, normalized=normalized,
uniform_noise=uniform_noise, noise=noise, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if centered is None:
centered = True
centered = _execute.make_bool(centered, "centered")
if normalized is None:
normalized = True
normalized = _execute.make_bool(normalized, "normalized")
if uniform_noise is None:
uniform_noise = True
uniform_noise = _execute.make_bool(uniform_noise, "uniform_noise")
if noise is None:
noise = "uniform"
noise = _execute.make_str(noise, "noise")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ExtractGlimpseV2", input=input, size=size, offsets=offsets,
centered=centered, normalized=normalized,
uniform_noise=uniform_noise, noise=noise,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("centered", _op._get_attr_bool("centered"), "normalized",
_op._get_attr_bool("normalized"), "uniform_noise",
_op._get_attr_bool("uniform_noise"), "noise",
_op.get_attr("noise"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ExtractGlimpseV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ExtractGlimpseV2 = tf_export("raw_ops.ExtractGlimpseV2")(_ops.to_raw_op(extract_glimpse_v2))
def extract_glimpse_v2_eager_fallback(input: Annotated[Any, _atypes.Float32], size: Annotated[Any, _atypes.Int32], offsets: Annotated[Any, _atypes.Float32], centered: bool, normalized: bool, uniform_noise: bool, noise: str, name, ctx) -> Annotated[Any, _atypes.Float32]:
if centered is None:
centered = True
centered = _execute.make_bool(centered, "centered")
if normalized is None:
normalized = True
normalized = _execute.make_bool(normalized, "normalized")
if uniform_noise is None:
uniform_noise = True
uniform_noise = _execute.make_bool(uniform_noise, "uniform_noise")
if noise is None:
noise = "uniform"
noise = _execute.make_str(noise, "noise")
input = _ops.convert_to_tensor(input, _dtypes.float32)
size = _ops.convert_to_tensor(size, _dtypes.int32)
offsets = _ops.convert_to_tensor(offsets, _dtypes.float32)
_inputs_flat = [input, size, offsets]
_attrs = ("centered", centered, "normalized", normalized, "uniform_noise",
uniform_noise, "noise", noise)
_result = _execute.execute(b"ExtractGlimpseV2", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ExtractGlimpseV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ExtractJpegShape_output_type = TypeVar("TV_ExtractJpegShape_output_type", _atypes.Int32, _atypes.Int64)
def extract_jpeg_shape(contents: Annotated[Any, _atypes.String], output_type:TV_ExtractJpegShape_output_type=_dtypes.int32, name=None) -> Annotated[Any, TV_ExtractJpegShape_output_type]:
r"""Extract the shape information of a JPEG-encoded image.
This op only parses the image header, so it is much faster than DecodeJpeg.
Args:
contents: A `Tensor` of type `string`. 0-D. The JPEG-encoded image.
output_type: An optional `tf.DType` from: `tf.int32, tf.int64`. Defaults to `tf.int32`.
(Optional) The output type of the operation (int32 or int64).
Defaults to int32.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `output_type`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ExtractJpegShape", name, contents, "output_type", output_type)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return extract_jpeg_shape_eager_fallback(
contents, output_type=output_type, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if output_type is None:
output_type = _dtypes.int32
output_type = _execute.make_type(output_type, "output_type")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ExtractJpegShape", contents=contents, output_type=output_type,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("output_type", _op._get_attr_type("output_type"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ExtractJpegShape", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ExtractJpegShape = tf_export("raw_ops.ExtractJpegShape")(_ops.to_raw_op(extract_jpeg_shape))
def extract_jpeg_shape_eager_fallback(contents: Annotated[Any, _atypes.String], output_type: TV_ExtractJpegShape_output_type, name, ctx) -> Annotated[Any, TV_ExtractJpegShape_output_type]:
if output_type is None:
output_type = _dtypes.int32
output_type = _execute.make_type(output_type, "output_type")
contents = _ops.convert_to_tensor(contents, _dtypes.string)
_inputs_flat = [contents]
_attrs = ("output_type", output_type)
_result = _execute.execute(b"ExtractJpegShape", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ExtractJpegShape", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_GenerateBoundingBoxProposalsOutput = collections.namedtuple(
"GenerateBoundingBoxProposals",
["rois", "roi_probabilities"])
def generate_bounding_box_proposals(scores: Annotated[Any, _atypes.Float32], bbox_deltas: Annotated[Any, _atypes.Float32], image_info: Annotated[Any, _atypes.Float32], anchors: Annotated[Any, _atypes.Float32], nms_threshold: Annotated[Any, _atypes.Float32], pre_nms_topn: Annotated[Any, _atypes.Int32], min_size: Annotated[Any, _atypes.Float32], post_nms_topn:int=300, name=None):
r"""This op produces Region of Interests from given bounding boxes(bbox_deltas) encoded wrt anchors according to eq.2 in arXiv:1506.01497
The op selects top `pre_nms_topn` scoring boxes, decodes them with respect to anchors,
applies non-maximal suppression on overlapping boxes with higher than
`nms_threshold` intersection-over-union (iou) value, discarding boxes where shorter
side is less than `min_size`.
Inputs:
`scores`: A 4D tensor of shape [Batch, Height, Width, Num Anchors] containing the scores per anchor at given position
`bbox_deltas`: is a tensor of shape [Batch, Height, Width, 4 x Num Anchors] boxes encoded to each anchor
`anchors`: A 1D tensor of shape [4 x Num Anchors], representing the anchors.
Outputs:
`rois`: output RoIs, a 3D tensor of shape [Batch, post_nms_topn, 4], padded by 0 if less than post_nms_topn candidates found.
`roi_probabilities`: probability scores of each roi in 'rois', a 2D tensor of shape [Batch,post_nms_topn], padded with 0 if needed, sorted by scores.
Args:
scores: A `Tensor` of type `float32`.
A 4-D float tensor of shape `[num_images, height, width, num_achors]` containing scores of the boxes for given anchors, can be unsorted.
bbox_deltas: A `Tensor` of type `float32`.
A 4-D float tensor of shape `[num_images, height, width, 4 x num_anchors]`. encoding boxes with respec to each anchor.
Coordinates are given in the form [dy, dx, dh, dw].
image_info: A `Tensor` of type `float32`.
A 2-D float tensor of shape `[num_images, 5]` containing image information Height, Width, Scale.
anchors: A `Tensor` of type `float32`.
A 2-D float tensor of shape `[num_anchors, 4]` describing the anchor boxes. Boxes are formatted in the form [y1, x1, y2, x2].
nms_threshold: A `Tensor` of type `float32`.
A scalar float tensor for non-maximal-suppression threshold.
pre_nms_topn: A `Tensor` of type `int32`.
A scalar int tensor for the number of top scoring boxes to be used as input.
min_size: A `Tensor` of type `float32`.
A scalar float tensor. Any box that has a smaller size than min_size will be discarded.
post_nms_topn: An optional `int`. Defaults to `300`.
An integer. Maximum number of rois in the output.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (rois, roi_probabilities).
rois: A `Tensor` of type `float32`.
roi_probabilities: A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "GenerateBoundingBoxProposals", name, scores, bbox_deltas,
image_info, anchors, nms_threshold, pre_nms_topn, min_size,
"post_nms_topn", post_nms_topn)
_result = _GenerateBoundingBoxProposalsOutput._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return generate_bounding_box_proposals_eager_fallback(
scores, bbox_deltas, image_info, anchors, nms_threshold,
pre_nms_topn, min_size, post_nms_topn=post_nms_topn, name=name,
ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if post_nms_topn is None:
post_nms_topn = 300
post_nms_topn = _execute.make_int(post_nms_topn, "post_nms_topn")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"GenerateBoundingBoxProposals", scores=scores,
bbox_deltas=bbox_deltas,
image_info=image_info,
anchors=anchors,
nms_threshold=nms_threshold,
pre_nms_topn=pre_nms_topn,
min_size=min_size,
post_nms_topn=post_nms_topn,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("post_nms_topn", _op._get_attr_int("post_nms_topn"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"GenerateBoundingBoxProposals", _inputs_flat, _attrs, _result)
_result = _GenerateBoundingBoxProposalsOutput._make(_result)
return _result
GenerateBoundingBoxProposals = tf_export("raw_ops.GenerateBoundingBoxProposals")(_ops.to_raw_op(generate_bounding_box_proposals))
def generate_bounding_box_proposals_eager_fallback(scores: Annotated[Any, _atypes.Float32], bbox_deltas: Annotated[Any, _atypes.Float32], image_info: Annotated[Any, _atypes.Float32], anchors: Annotated[Any, _atypes.Float32], nms_threshold: Annotated[Any, _atypes.Float32], pre_nms_topn: Annotated[Any, _atypes.Int32], min_size: Annotated[Any, _atypes.Float32], post_nms_topn: int, name, ctx):
if post_nms_topn is None:
post_nms_topn = 300
post_nms_topn = _execute.make_int(post_nms_topn, "post_nms_topn")
scores = _ops.convert_to_tensor(scores, _dtypes.float32)
bbox_deltas = _ops.convert_to_tensor(bbox_deltas, _dtypes.float32)
image_info = _ops.convert_to_tensor(image_info, _dtypes.float32)
anchors = _ops.convert_to_tensor(anchors, _dtypes.float32)
nms_threshold = _ops.convert_to_tensor(nms_threshold, _dtypes.float32)
pre_nms_topn = _ops.convert_to_tensor(pre_nms_topn, _dtypes.int32)
min_size = _ops.convert_to_tensor(min_size, _dtypes.float32)
_inputs_flat = [scores, bbox_deltas, image_info, anchors, nms_threshold, pre_nms_topn, min_size]
_attrs = ("post_nms_topn", post_nms_topn)
_result = _execute.execute(b"GenerateBoundingBoxProposals", 2,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"GenerateBoundingBoxProposals", _inputs_flat, _attrs, _result)
_result = _GenerateBoundingBoxProposalsOutput._make(_result)
return _result
TV_HSVToRGB_T = TypeVar("TV_HSVToRGB_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half)
@_dispatch.add_fallback_dispatch_list
@_dispatch.add_type_based_api_dispatcher
@tf_export('image.hsv_to_rgb')
def hsv_to_rgb(images: Annotated[Any, TV_HSVToRGB_T], name=None) -> Annotated[Any, TV_HSVToRGB_T]:
r"""Convert one or more images from HSV to RGB.
Outputs a tensor of the same shape as the `images` tensor, containing the RGB
value of the pixels. The output is only well defined if the value in `images`
are in `[0,1]`.
See `rgb_to_hsv` for a description of the HSV encoding.
Args:
images: A `Tensor`. Must be one of the following types: `half`, `bfloat16`, `float32`, `float64`.
1-D or higher rank. HSV data to convert. Last dimension must be size 3.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "HSVToRGB", name, images)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
_result = _dispatcher_for_hsv_to_rgb(
(images, name,), None)
if _result is not NotImplemented:
return _result
return hsv_to_rgb_eager_fallback(
images, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
except (TypeError, ValueError):
_result = _dispatch.dispatch(
hsv_to_rgb, (), dict(images=images, name=name)
)
if _result is not _dispatch.OpDispatcher.NOT_SUPPORTED:
return _result
raise
else:
_result = _dispatcher_for_hsv_to_rgb(
(images, name,), None)
if _result is not NotImplemented:
return _result
# Add nodes to the TensorFlow graph.
try:
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"HSVToRGB", images=images, name=name)
except (TypeError, ValueError):
_result = _dispatch.dispatch(
hsv_to_rgb, (), dict(images=images, name=name)
)
if _result is not _dispatch.OpDispatcher.NOT_SUPPORTED:
return _result
raise
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"HSVToRGB", _inputs_flat, _attrs, _result)
_result, = _result
return _result
HSVToRGB = tf_export("raw_ops.HSVToRGB")(_ops.to_raw_op(hsv_to_rgb))
_dispatcher_for_hsv_to_rgb = hsv_to_rgb._tf_type_based_dispatcher.Dispatch
def hsv_to_rgb_eager_fallback(images: Annotated[Any, TV_HSVToRGB_T], name, ctx) -> Annotated[Any, TV_HSVToRGB_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.half, _dtypes.bfloat16, _dtypes.float32, _dtypes.float64, ], _dtypes.float32)
_inputs_flat = [images]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"HSVToRGB", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"HSVToRGB", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ImageProjectiveTransformV2_dtype = TypeVar("TV_ImageProjectiveTransformV2_dtype", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int32, _atypes.Int64, _atypes.UInt8)
def image_projective_transform_v2(images: Annotated[Any, TV_ImageProjectiveTransformV2_dtype], transforms: Annotated[Any, _atypes.Float32], output_shape: Annotated[Any, _atypes.Int32], interpolation: str, fill_mode:str="CONSTANT", name=None) -> Annotated[Any, TV_ImageProjectiveTransformV2_dtype]:
r"""Applies the given transform to each of the images.
If one row of `transforms` is `[a0, a1, a2, b0, b1, b2, c0, c1]`, then it maps
the *output* point `(x, y)` to a transformed *input* point
`(x', y') = ((a0 x + a1 y + a2) / k, (b0 x + b1 y + b2) / k)`, where
`k = c0 x + c1 y + 1`. If the transformed point lays outside of the input
image, the output pixel is set to 0.
Args:
images: A `Tensor`. Must be one of the following types: `uint8`, `int32`, `int64`, `half`, `bfloat16`, `float32`, `float64`.
4-D with shape `[batch, height, width, channels]`.
transforms: A `Tensor` of type `float32`.
2-D Tensor, `[batch, 8]` or `[1, 8]` matrix, where each row corresponds to a 3 x 3
projective transformation matrix, with the last entry assumed to be 1. If there
is one row, the same transformation will be applied to all images.
output_shape: A `Tensor` of type `int32`.
1-D Tensor [new_height, new_width].
interpolation: A `string`. Interpolation method, "NEAREST" or "BILINEAR".
fill_mode: An optional `string`. Defaults to `"CONSTANT"`.
Fill mode, "REFLECT", "WRAP", or "CONSTANT".
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ImageProjectiveTransformV2", name, images, transforms,
output_shape, "interpolation", interpolation, "fill_mode", fill_mode)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return image_projective_transform_v2_eager_fallback(
images, transforms, output_shape, interpolation=interpolation,
fill_mode=fill_mode, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
interpolation = _execute.make_str(interpolation, "interpolation")
if fill_mode is None:
fill_mode = "CONSTANT"
fill_mode = _execute.make_str(fill_mode, "fill_mode")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ImageProjectiveTransformV2", images=images, transforms=transforms,
output_shape=output_shape,
interpolation=interpolation,
fill_mode=fill_mode, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("dtype", _op._get_attr_type("dtype"), "interpolation",
_op.get_attr("interpolation"), "fill_mode",
_op.get_attr("fill_mode"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ImageProjectiveTransformV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ImageProjectiveTransformV2 = tf_export("raw_ops.ImageProjectiveTransformV2")(_ops.to_raw_op(image_projective_transform_v2))
def image_projective_transform_v2_eager_fallback(images: Annotated[Any, TV_ImageProjectiveTransformV2_dtype], transforms: Annotated[Any, _atypes.Float32], output_shape: Annotated[Any, _atypes.Int32], interpolation: str, fill_mode: str, name, ctx) -> Annotated[Any, TV_ImageProjectiveTransformV2_dtype]:
interpolation = _execute.make_str(interpolation, "interpolation")
if fill_mode is None:
fill_mode = "CONSTANT"
fill_mode = _execute.make_str(fill_mode, "fill_mode")
_attr_dtype, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.uint8, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.bfloat16, _dtypes.float32, _dtypes.float64, ])
transforms = _ops.convert_to_tensor(transforms, _dtypes.float32)
output_shape = _ops.convert_to_tensor(output_shape, _dtypes.int32)
_inputs_flat = [images, transforms, output_shape]
_attrs = ("dtype", _attr_dtype, "interpolation", interpolation, "fill_mode",
fill_mode)
_result = _execute.execute(b"ImageProjectiveTransformV2", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ImageProjectiveTransformV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ImageProjectiveTransformV3_dtype = TypeVar("TV_ImageProjectiveTransformV3_dtype", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int32, _atypes.Int64, _atypes.UInt8)
def image_projective_transform_v3(images: Annotated[Any, TV_ImageProjectiveTransformV3_dtype], transforms: Annotated[Any, _atypes.Float32], output_shape: Annotated[Any, _atypes.Int32], fill_value: Annotated[Any, _atypes.Float32], interpolation: str, fill_mode:str="CONSTANT", name=None) -> Annotated[Any, TV_ImageProjectiveTransformV3_dtype]:
r"""Applies the given transform to each of the images.
If one row of `transforms` is `[a0, a1, a2, b0, b1, b2, c0, c1]`, then it maps
the *output* point `(x, y)` to a transformed *input* point
`(x', y') = ((a0 x + a1 y + a2) / k, (b0 x + b1 y + b2) / k)`, where
`k = c0 x + c1 y + 1`. If the transformed point lays outside of the input
image, the output pixel is set to fill_value.
Args:
images: A `Tensor`. Must be one of the following types: `uint8`, `int32`, `int64`, `half`, `bfloat16`, `float32`, `float64`.
4-D with shape `[batch, height, width, channels]`.
transforms: A `Tensor` of type `float32`.
2-D Tensor, `[batch, 8]` or `[1, 8]` matrix, where each row corresponds to a 3 x 3
projective transformation matrix, with the last entry assumed to be 1. If there
is one row, the same transformation will be applied to all images.
output_shape: A `Tensor` of type `int32`.
1-D Tensor [new_height, new_width].
fill_value: A `Tensor` of type `float32`.
float, the value to be filled when fill_mode is constant".
END
}
out_arg {
name: "transformed_images"
description: <<END
4-D with shape
`[batch, new_height, new_width, channels]`.
interpolation: A `string`. Interpolation method, "NEAREST" or "BILINEAR".
fill_mode: An optional `string`. Defaults to `"CONSTANT"`.
Fill mode, "REFLECT", "WRAP", "CONSTANT", or "NEAREST".
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ImageProjectiveTransformV3", name, images, transforms,
output_shape, fill_value, "interpolation", interpolation, "fill_mode",
fill_mode)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return image_projective_transform_v3_eager_fallback(
images, transforms, output_shape, fill_value,
interpolation=interpolation, fill_mode=fill_mode, name=name,
ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
interpolation = _execute.make_str(interpolation, "interpolation")
if fill_mode is None:
fill_mode = "CONSTANT"
fill_mode = _execute.make_str(fill_mode, "fill_mode")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ImageProjectiveTransformV3", images=images, transforms=transforms,
output_shape=output_shape,
fill_value=fill_value,
interpolation=interpolation,
fill_mode=fill_mode, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("dtype", _op._get_attr_type("dtype"), "interpolation",
_op.get_attr("interpolation"), "fill_mode",
_op.get_attr("fill_mode"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ImageProjectiveTransformV3", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ImageProjectiveTransformV3 = tf_export("raw_ops.ImageProjectiveTransformV3")(_ops.to_raw_op(image_projective_transform_v3))
def image_projective_transform_v3_eager_fallback(images: Annotated[Any, TV_ImageProjectiveTransformV3_dtype], transforms: Annotated[Any, _atypes.Float32], output_shape: Annotated[Any, _atypes.Int32], fill_value: Annotated[Any, _atypes.Float32], interpolation: str, fill_mode: str, name, ctx) -> Annotated[Any, TV_ImageProjectiveTransformV3_dtype]:
interpolation = _execute.make_str(interpolation, "interpolation")
if fill_mode is None:
fill_mode = "CONSTANT"
fill_mode = _execute.make_str(fill_mode, "fill_mode")
_attr_dtype, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.uint8, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.bfloat16, _dtypes.float32, _dtypes.float64, ])
transforms = _ops.convert_to_tensor(transforms, _dtypes.float32)
output_shape = _ops.convert_to_tensor(output_shape, _dtypes.int32)
fill_value = _ops.convert_to_tensor(fill_value, _dtypes.float32)
_inputs_flat = [images, transforms, output_shape, fill_value]
_attrs = ("dtype", _attr_dtype, "interpolation", interpolation, "fill_mode",
fill_mode)
_result = _execute.execute(b"ImageProjectiveTransformV3", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ImageProjectiveTransformV3", _inputs_flat, _attrs, _result)
_result, = _result
return _result
def non_max_suppression(boxes: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold:float=0.5, name=None) -> Annotated[Any, _atypes.Int32]:
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system. Note that this
algorithm is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression(
boxes, scores, max_output_size, iou_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Args:
boxes: A `Tensor` of type `float32`.
A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A `Tensor` of type `float32`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
iou_threshold: An optional `float`. Defaults to `0.5`.
A float representing the threshold for deciding whether boxes
overlap too much with respect to IOU.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppression", name, boxes, scores, max_output_size,
"iou_threshold", iou_threshold)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_eager_fallback(
boxes, scores, max_output_size, iou_threshold=iou_threshold,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if iou_threshold is None:
iou_threshold = 0.5
iou_threshold = _execute.make_float(iou_threshold, "iou_threshold")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppression", boxes=boxes, scores=scores,
max_output_size=max_output_size,
iou_threshold=iou_threshold, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("iou_threshold", _op.get_attr("iou_threshold"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppression", _inputs_flat, _attrs, _result)
_result, = _result
return _result
NonMaxSuppression = tf_export("raw_ops.NonMaxSuppression")(_ops.to_raw_op(non_max_suppression))
def non_max_suppression_eager_fallback(boxes: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: float, name, ctx) -> Annotated[Any, _atypes.Int32]:
if iou_threshold is None:
iou_threshold = 0.5
iou_threshold = _execute.make_float(iou_threshold, "iou_threshold")
boxes = _ops.convert_to_tensor(boxes, _dtypes.float32)
scores = _ops.convert_to_tensor(scores, _dtypes.float32)
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
_inputs_flat = [boxes, scores, max_output_size]
_attrs = ("iou_threshold", iou_threshold)
_result = _execute.execute(b"NonMaxSuppression", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppression", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_NonMaxSuppressionV2_T = TypeVar("TV_NonMaxSuppressionV2_T", _atypes.Float32, _atypes.Half)
TV_NonMaxSuppressionV2_T_threshold = TypeVar("TV_NonMaxSuppressionV2_T_threshold", _atypes.Float32, _atypes.Half)
def non_max_suppression_v2(boxes: Annotated[Any, TV_NonMaxSuppressionV2_T], scores: Annotated[Any, TV_NonMaxSuppressionV2_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV2_T_threshold], name=None) -> Annotated[Any, _atypes.Int32]:
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system. Note that this
algorithm is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression_v2(
boxes, scores, max_output_size, iou_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Args:
boxes: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A `Tensor`. Must have the same type as `boxes`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
iou_threshold: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too much with respect to IOU.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppressionV2", name, boxes, scores, max_output_size,
iou_threshold)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_v2_eager_fallback(
boxes, scores, max_output_size, iou_threshold, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppressionV2", boxes=boxes, scores=scores,
max_output_size=max_output_size,
iou_threshold=iou_threshold, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "T_threshold",
_op._get_attr_type("T_threshold"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppressionV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
NonMaxSuppressionV2 = tf_export("raw_ops.NonMaxSuppressionV2")(_ops.to_raw_op(non_max_suppression_v2))
def non_max_suppression_v2_eager_fallback(boxes: Annotated[Any, TV_NonMaxSuppressionV2_T], scores: Annotated[Any, TV_NonMaxSuppressionV2_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV2_T_threshold], name, ctx) -> Annotated[Any, _atypes.Int32]:
_attr_T, _inputs_T = _execute.args_to_matching_eager([boxes, scores], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(boxes, scores) = _inputs_T
_attr_T_threshold, (iou_threshold,) = _execute.args_to_matching_eager([iou_threshold], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
_inputs_flat = [boxes, scores, max_output_size, iou_threshold]
_attrs = ("T", _attr_T, "T_threshold", _attr_T_threshold)
_result = _execute.execute(b"NonMaxSuppressionV2", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppressionV2", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_NonMaxSuppressionV3_T = TypeVar("TV_NonMaxSuppressionV3_T", _atypes.Float32, _atypes.Half)
TV_NonMaxSuppressionV3_T_threshold = TypeVar("TV_NonMaxSuppressionV3_T_threshold", _atypes.Float32, _atypes.Half)
def non_max_suppression_v3(boxes: Annotated[Any, TV_NonMaxSuppressionV3_T], scores: Annotated[Any, TV_NonMaxSuppressionV3_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV3_T_threshold], score_threshold: Annotated[Any, TV_NonMaxSuppressionV3_T_threshold], name=None) -> Annotated[Any, _atypes.Int32]:
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes with score less than
`score_threshold` are removed. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system and more
generally is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression_v2(
boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Args:
boxes: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A `Tensor`. Must have the same type as `boxes`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
iou_threshold: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too much with respect to IOU.
score_threshold: A `Tensor`. Must have the same type as `iou_threshold`.
A 0-D float tensor representing the threshold for deciding when to remove
boxes based on score.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppressionV3", name, boxes, scores, max_output_size,
iou_threshold, score_threshold)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_v3_eager_fallback(
boxes, scores, max_output_size, iou_threshold, score_threshold,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppressionV3", boxes=boxes, scores=scores,
max_output_size=max_output_size,
iou_threshold=iou_threshold,
score_threshold=score_threshold, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "T_threshold",
_op._get_attr_type("T_threshold"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppressionV3", _inputs_flat, _attrs, _result)
_result, = _result
return _result
NonMaxSuppressionV3 = tf_export("raw_ops.NonMaxSuppressionV3")(_ops.to_raw_op(non_max_suppression_v3))
def non_max_suppression_v3_eager_fallback(boxes: Annotated[Any, TV_NonMaxSuppressionV3_T], scores: Annotated[Any, TV_NonMaxSuppressionV3_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV3_T_threshold], score_threshold: Annotated[Any, TV_NonMaxSuppressionV3_T_threshold], name, ctx) -> Annotated[Any, _atypes.Int32]:
_attr_T, _inputs_T = _execute.args_to_matching_eager([boxes, scores], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(boxes, scores) = _inputs_T
_attr_T_threshold, _inputs_T_threshold = _execute.args_to_matching_eager([iou_threshold, score_threshold], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(iou_threshold, score_threshold) = _inputs_T_threshold
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
_inputs_flat = [boxes, scores, max_output_size, iou_threshold, score_threshold]
_attrs = ("T", _attr_T, "T_threshold", _attr_T_threshold)
_result = _execute.execute(b"NonMaxSuppressionV3", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppressionV3", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_NonMaxSuppressionV4Output = collections.namedtuple(
"NonMaxSuppressionV4",
["selected_indices", "valid_outputs"])
TV_NonMaxSuppressionV4_T = TypeVar("TV_NonMaxSuppressionV4_T", _atypes.Float32, _atypes.Half)
TV_NonMaxSuppressionV4_T_threshold = TypeVar("TV_NonMaxSuppressionV4_T_threshold", _atypes.Float32, _atypes.Half)
def non_max_suppression_v4(boxes: Annotated[Any, TV_NonMaxSuppressionV4_T], scores: Annotated[Any, TV_NonMaxSuppressionV4_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV4_T_threshold], score_threshold: Annotated[Any, TV_NonMaxSuppressionV4_T_threshold], pad_to_max_output_size:bool=False, name=None):
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes with score less than
`score_threshold` are removed. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system and more
generally is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression_v2(
boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Args:
boxes: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A `Tensor`. Must have the same type as `boxes`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
iou_threshold: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too much with respect to IOU.
score_threshold: A `Tensor`. Must have the same type as `iou_threshold`.
A 0-D float tensor representing the threshold for deciding when to remove
boxes based on score.
pad_to_max_output_size: An optional `bool`. Defaults to `False`.
If true, the output `selected_indices` is padded to be of length
`max_output_size`. Defaults to false.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (selected_indices, valid_outputs).
selected_indices: A `Tensor` of type `int32`.
valid_outputs: A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppressionV4", name, boxes, scores, max_output_size,
iou_threshold, score_threshold, "pad_to_max_output_size",
pad_to_max_output_size)
_result = _NonMaxSuppressionV4Output._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_v4_eager_fallback(
boxes, scores, max_output_size, iou_threshold, score_threshold,
pad_to_max_output_size=pad_to_max_output_size, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if pad_to_max_output_size is None:
pad_to_max_output_size = False
pad_to_max_output_size = _execute.make_bool(pad_to_max_output_size, "pad_to_max_output_size")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppressionV4", boxes=boxes, scores=scores,
max_output_size=max_output_size,
iou_threshold=iou_threshold,
score_threshold=score_threshold,
pad_to_max_output_size=pad_to_max_output_size,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "T_threshold",
_op._get_attr_type("T_threshold"), "pad_to_max_output_size",
_op._get_attr_bool("pad_to_max_output_size"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppressionV4", _inputs_flat, _attrs, _result)
_result = _NonMaxSuppressionV4Output._make(_result)
return _result
NonMaxSuppressionV4 = tf_export("raw_ops.NonMaxSuppressionV4")(_ops.to_raw_op(non_max_suppression_v4))
def non_max_suppression_v4_eager_fallback(boxes: Annotated[Any, TV_NonMaxSuppressionV4_T], scores: Annotated[Any, TV_NonMaxSuppressionV4_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV4_T_threshold], score_threshold: Annotated[Any, TV_NonMaxSuppressionV4_T_threshold], pad_to_max_output_size: bool, name, ctx):
if pad_to_max_output_size is None:
pad_to_max_output_size = False
pad_to_max_output_size = _execute.make_bool(pad_to_max_output_size, "pad_to_max_output_size")
_attr_T, _inputs_T = _execute.args_to_matching_eager([boxes, scores], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(boxes, scores) = _inputs_T
_attr_T_threshold, _inputs_T_threshold = _execute.args_to_matching_eager([iou_threshold, score_threshold], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(iou_threshold, score_threshold) = _inputs_T_threshold
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
_inputs_flat = [boxes, scores, max_output_size, iou_threshold, score_threshold]
_attrs = ("T", _attr_T, "T_threshold", _attr_T_threshold,
"pad_to_max_output_size", pad_to_max_output_size)
_result = _execute.execute(b"NonMaxSuppressionV4", 2, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppressionV4", _inputs_flat, _attrs, _result)
_result = _NonMaxSuppressionV4Output._make(_result)
return _result
_NonMaxSuppressionV5Output = collections.namedtuple(
"NonMaxSuppressionV5",
["selected_indices", "selected_scores", "valid_outputs"])
TV_NonMaxSuppressionV5_T = TypeVar("TV_NonMaxSuppressionV5_T", _atypes.Float32, _atypes.Half)
def non_max_suppression_v5(boxes: Annotated[Any, TV_NonMaxSuppressionV5_T], scores: Annotated[Any, TV_NonMaxSuppressionV5_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV5_T], score_threshold: Annotated[Any, TV_NonMaxSuppressionV5_T], soft_nms_sigma: Annotated[Any, TV_NonMaxSuppressionV5_T], pad_to_max_output_size:bool=False, name=None):
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high intersection-over-union (IOU) overlap
with previously selected boxes. Bounding boxes with score less than
`score_threshold` are removed. Bounding boxes are supplied as
[y1, x1, y2, x2], where (y1, x1) and (y2, x2) are the coordinates of any
diagonal pair of box corners and the coordinates can be provided as normalized
(i.e., lying in the interval [0, 1]) or absolute. Note that this algorithm
is agnostic to where the origin is in the coordinate system and more
generally is invariant to orthogonal transformations and translations
of the coordinate system; thus translating or reflections of the coordinate
system result in the same boxes being selected by the algorithm.
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression_v2(
boxes, scores, max_output_size, iou_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
This op also supports a Soft-NMS (with Gaussian weighting) mode (c.f.
Bodla et al, https://arxiv.org/abs/1704.04503) where boxes reduce the score
of other overlapping boxes instead of directly causing them to be pruned.
To enable this Soft-NMS mode, set the `soft_nms_sigma` parameter to be
larger than 0.
Args:
boxes: A `Tensor`. Must be one of the following types: `half`, `float32`.
A 2-D float tensor of shape `[num_boxes, 4]`.
scores: A `Tensor`. Must have the same type as `boxes`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
iou_threshold: A `Tensor`. Must have the same type as `boxes`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too much with respect to IOU.
score_threshold: A `Tensor`. Must have the same type as `boxes`.
A 0-D float tensor representing the threshold for deciding when to remove
boxes based on score.
soft_nms_sigma: A `Tensor`. Must have the same type as `boxes`.
A 0-D float tensor representing the sigma parameter for Soft NMS; see Bodla et
al (c.f. https://arxiv.org/abs/1704.04503). When `soft_nms_sigma=0.0` (which
is default), we fall back to standard (hard) NMS.
pad_to_max_output_size: An optional `bool`. Defaults to `False`.
If true, the output `selected_indices` is padded to be of length
`max_output_size`. Defaults to false.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (selected_indices, selected_scores, valid_outputs).
selected_indices: A `Tensor` of type `int32`.
selected_scores: A `Tensor`. Has the same type as `boxes`.
valid_outputs: A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppressionV5", name, boxes, scores, max_output_size,
iou_threshold, score_threshold, soft_nms_sigma,
"pad_to_max_output_size", pad_to_max_output_size)
_result = _NonMaxSuppressionV5Output._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_v5_eager_fallback(
boxes, scores, max_output_size, iou_threshold, score_threshold,
soft_nms_sigma, pad_to_max_output_size=pad_to_max_output_size,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if pad_to_max_output_size is None:
pad_to_max_output_size = False
pad_to_max_output_size = _execute.make_bool(pad_to_max_output_size, "pad_to_max_output_size")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppressionV5", boxes=boxes, scores=scores,
max_output_size=max_output_size,
iou_threshold=iou_threshold,
score_threshold=score_threshold,
soft_nms_sigma=soft_nms_sigma,
pad_to_max_output_size=pad_to_max_output_size,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "pad_to_max_output_size",
_op._get_attr_bool("pad_to_max_output_size"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppressionV5", _inputs_flat, _attrs, _result)
_result = _NonMaxSuppressionV5Output._make(_result)
return _result
NonMaxSuppressionV5 = tf_export("raw_ops.NonMaxSuppressionV5")(_ops.to_raw_op(non_max_suppression_v5))
def non_max_suppression_v5_eager_fallback(boxes: Annotated[Any, TV_NonMaxSuppressionV5_T], scores: Annotated[Any, TV_NonMaxSuppressionV5_T], max_output_size: Annotated[Any, _atypes.Int32], iou_threshold: Annotated[Any, TV_NonMaxSuppressionV5_T], score_threshold: Annotated[Any, TV_NonMaxSuppressionV5_T], soft_nms_sigma: Annotated[Any, TV_NonMaxSuppressionV5_T], pad_to_max_output_size: bool, name, ctx):
if pad_to_max_output_size is None:
pad_to_max_output_size = False
pad_to_max_output_size = _execute.make_bool(pad_to_max_output_size, "pad_to_max_output_size")
_attr_T, _inputs_T = _execute.args_to_matching_eager([boxes, scores, iou_threshold, score_threshold, soft_nms_sigma], ctx, [_dtypes.half, _dtypes.float32, ], _dtypes.float32)
(boxes, scores, iou_threshold, score_threshold, soft_nms_sigma) = _inputs_T
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
_inputs_flat = [boxes, scores, max_output_size, iou_threshold, score_threshold, soft_nms_sigma]
_attrs = ("T", _attr_T, "pad_to_max_output_size", pad_to_max_output_size)
_result = _execute.execute(b"NonMaxSuppressionV5", 3, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppressionV5", _inputs_flat, _attrs, _result)
_result = _NonMaxSuppressionV5Output._make(_result)
return _result
def non_max_suppression_with_overlaps(overlaps: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size: Annotated[Any, _atypes.Int32], overlap_threshold: Annotated[Any, _atypes.Float32], score_threshold: Annotated[Any, _atypes.Float32], name=None) -> Annotated[Any, _atypes.Int32]:
r"""Greedily selects a subset of bounding boxes in descending order of score,
pruning away boxes that have high overlaps
with previously selected boxes. Bounding boxes with score less than
`score_threshold` are removed. N-by-n overlap values are supplied as square matrix,
which allows for defining a custom overlap criterium (eg. intersection over union,
intersection over area, etc.).
The output of this operation is a set of integers indexing into the input
collection of bounding boxes representing the selected boxes. The bounding
box coordinates corresponding to the selected indices can then be obtained
using the `tf.gather operation`. For example:
selected_indices = tf.image.non_max_suppression_with_overlaps(
overlaps, scores, max_output_size, overlap_threshold, score_threshold)
selected_boxes = tf.gather(boxes, selected_indices)
Args:
overlaps: A `Tensor` of type `float32`.
A 2-D float tensor of shape `[num_boxes, num_boxes]` representing
the n-by-n box overlap values.
scores: A `Tensor` of type `float32`.
A 1-D float tensor of shape `[num_boxes]` representing a single
score corresponding to each box (each row of boxes).
max_output_size: A `Tensor` of type `int32`.
A scalar integer tensor representing the maximum number of
boxes to be selected by non max suppression.
overlap_threshold: A `Tensor` of type `float32`.
A 0-D float tensor representing the threshold for deciding whether
boxes overlap too.
score_threshold: A `Tensor` of type `float32`.
A 0-D float tensor representing the threshold for deciding when to remove
boxes based on score.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `int32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "NonMaxSuppressionWithOverlaps", name, overlaps, scores,
max_output_size, overlap_threshold, score_threshold)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return non_max_suppression_with_overlaps_eager_fallback(
overlaps, scores, max_output_size, overlap_threshold,
score_threshold, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"NonMaxSuppressionWithOverlaps", overlaps=overlaps, scores=scores,
max_output_size=max_output_size,
overlap_threshold=overlap_threshold,
score_threshold=score_threshold,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ()
_inputs_flat = _op.inputs
_execute.record_gradient(
"NonMaxSuppressionWithOverlaps", _inputs_flat, _attrs, _result)
_result, = _result
return _result
NonMaxSuppressionWithOverlaps = tf_export("raw_ops.NonMaxSuppressionWithOverlaps")(_ops.to_raw_op(non_max_suppression_with_overlaps))
def non_max_suppression_with_overlaps_eager_fallback(overlaps: Annotated[Any, _atypes.Float32], scores: Annotated[Any, _atypes.Float32], max_output_size: Annotated[Any, _atypes.Int32], overlap_threshold: Annotated[Any, _atypes.Float32], score_threshold: Annotated[Any, _atypes.Float32], name, ctx) -> Annotated[Any, _atypes.Int32]:
overlaps = _ops.convert_to_tensor(overlaps, _dtypes.float32)
scores = _ops.convert_to_tensor(scores, _dtypes.float32)
max_output_size = _ops.convert_to_tensor(max_output_size, _dtypes.int32)
overlap_threshold = _ops.convert_to_tensor(overlap_threshold, _dtypes.float32)
score_threshold = _ops.convert_to_tensor(score_threshold, _dtypes.float32)
_inputs_flat = [overlaps, scores, max_output_size, overlap_threshold, score_threshold]
_attrs = None
_result = _execute.execute(b"NonMaxSuppressionWithOverlaps", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"NonMaxSuppressionWithOverlaps", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_QuantizedResizeBilinearOutput = collections.namedtuple(
"QuantizedResizeBilinear",
["resized_images", "out_min", "out_max"])
TV_QuantizedResizeBilinear_T = TypeVar("TV_QuantizedResizeBilinear_T", _atypes.Float32, _atypes.QInt32, _atypes.QUInt8)
def quantized_resize_bilinear(images: Annotated[Any, TV_QuantizedResizeBilinear_T], size: Annotated[Any, _atypes.Int32], min: Annotated[Any, _atypes.Float32], max: Annotated[Any, _atypes.Float32], align_corners:bool=False, half_pixel_centers:bool=False, name=None):
r"""Resize quantized `images` to `size` using quantized bilinear interpolation.
Input images and output images must be quantized types.
Args:
images: A `Tensor`. Must be one of the following types: `quint8`, `qint32`, `float32`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
min: A `Tensor` of type `float32`.
max: A `Tensor` of type `float32`.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and output tensors are
aligned, preserving the values at the corner pixels. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (resized_images, out_min, out_max).
resized_images: A `Tensor`. Has the same type as `images`.
out_min: A `Tensor` of type `float32`.
out_max: A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "QuantizedResizeBilinear", name, images, size, min, max,
"align_corners", align_corners, "half_pixel_centers",
half_pixel_centers)
_result = _QuantizedResizeBilinearOutput._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return quantized_resize_bilinear_eager_fallback(
images, size, min, max, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"QuantizedResizeBilinear", images=images, size=size, min=min, max=max,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"QuantizedResizeBilinear", _inputs_flat, _attrs, _result)
_result = _QuantizedResizeBilinearOutput._make(_result)
return _result
QuantizedResizeBilinear = tf_export("raw_ops.QuantizedResizeBilinear")(_ops.to_raw_op(quantized_resize_bilinear))
def quantized_resize_bilinear_eager_fallback(images: Annotated[Any, TV_QuantizedResizeBilinear_T], size: Annotated[Any, _atypes.Int32], min: Annotated[Any, _atypes.Float32], max: Annotated[Any, _atypes.Float32], align_corners: bool, half_pixel_centers: bool, name, ctx):
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.quint8, _dtypes.qint32, _dtypes.float32, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
min = _ops.convert_to_tensor(min, _dtypes.float32)
max = _ops.convert_to_tensor(max, _dtypes.float32)
_inputs_flat = [images, size, min, max]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"QuantizedResizeBilinear", 3,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"QuantizedResizeBilinear", _inputs_flat, _attrs, _result)
_result = _QuantizedResizeBilinearOutput._make(_result)
return _result
TV_RGBToHSV_T = TypeVar("TV_RGBToHSV_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half)
@_dispatch.add_fallback_dispatch_list
@_dispatch.add_type_based_api_dispatcher
@tf_export('image.rgb_to_hsv')
def rgb_to_hsv(images: Annotated[Any, TV_RGBToHSV_T], name=None) -> Annotated[Any, TV_RGBToHSV_T]:
r"""Converts one or more images from RGB to HSV.
Outputs a tensor of the same shape as the `images` tensor, containing the HSV
value of the pixels. The output is only well defined if the value in `images`
are in `[0,1]`.
`output[..., 0]` contains hue, `output[..., 1]` contains saturation, and
`output[..., 2]` contains value. All HSV values are in `[0,1]`. A hue of 0
corresponds to pure red, hue 1/3 is pure green, and 2/3 is pure blue.
Usage Example:
>>> blue_image = tf.stack([
... tf.zeros([5,5]),
... tf.zeros([5,5]),
... tf.ones([5,5])],
... axis=-1)
>>> blue_hsv_image = tf.image.rgb_to_hsv(blue_image)
>>> blue_hsv_image[0,0].numpy()
array([0.6666667, 1. , 1. ], dtype=float32)
Args:
images: A `Tensor`. Must be one of the following types: `half`, `bfloat16`, `float32`, `float64`.
1-D or higher rank. RGB data to convert. Last dimension must be size 3.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "RGBToHSV", name, images)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
_result = _dispatcher_for_rgb_to_hsv(
(images, name,), None)
if _result is not NotImplemented:
return _result
return rgb_to_hsv_eager_fallback(
images, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
except (TypeError, ValueError):
_result = _dispatch.dispatch(
rgb_to_hsv, (), dict(images=images, name=name)
)
if _result is not _dispatch.OpDispatcher.NOT_SUPPORTED:
return _result
raise
else:
_result = _dispatcher_for_rgb_to_hsv(
(images, name,), None)
if _result is not NotImplemented:
return _result
# Add nodes to the TensorFlow graph.
try:
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"RGBToHSV", images=images, name=name)
except (TypeError, ValueError):
_result = _dispatch.dispatch(
rgb_to_hsv, (), dict(images=images, name=name)
)
if _result is not _dispatch.OpDispatcher.NOT_SUPPORTED:
return _result
raise
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"RGBToHSV", _inputs_flat, _attrs, _result)
_result, = _result
return _result
RGBToHSV = tf_export("raw_ops.RGBToHSV")(_ops.to_raw_op(rgb_to_hsv))
_dispatcher_for_rgb_to_hsv = rgb_to_hsv._tf_type_based_dispatcher.Dispatch
def rgb_to_hsv_eager_fallback(images: Annotated[Any, TV_RGBToHSV_T], name, ctx) -> Annotated[Any, TV_RGBToHSV_T]:
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.half, _dtypes.bfloat16, _dtypes.float32, _dtypes.float64, ], _dtypes.float32)
_inputs_flat = [images]
_attrs = ("T", _attr_T)
_result = _execute.execute(b"RGBToHSV", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"RGBToHSV", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_RandomCrop_T = TypeVar("TV_RandomCrop_T", _atypes.Float32, _atypes.Float64, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt8)
def random_crop(image: Annotated[Any, TV_RandomCrop_T], size: Annotated[Any, _atypes.Int64], seed:int=0, seed2:int=0, name=None) -> Annotated[Any, TV_RandomCrop_T]:
r"""Randomly crop `image`.
`size` is a 1-D int64 tensor with 2 elements representing the crop height and
width. The values must be non negative.
This Op picks a random location in `image` and crops a `height` by `width`
rectangle from that location. The random location is picked so the cropped
area will fit inside the original image.
Args:
image: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int16`, `int32`, `int64`, `float32`, `float64`.
3-D of shape `[height, width, channels]`.
size: A `Tensor` of type `int64`.
1-D of length 2 containing: `crop_height`, `crop_width`..
seed: An optional `int`. Defaults to `0`.
If either seed or seed2 are set to be non-zero, the random number
generator is seeded by the given seed. Otherwise, it is seeded by a
random seed.
seed2: An optional `int`. Defaults to `0`.
An second seed to avoid seed collision.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `image`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "RandomCrop", name, image, size, "seed", seed, "seed2", seed2)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return random_crop_eager_fallback(
image, size, seed=seed, seed2=seed2, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"RandomCrop", image=image, size=size, seed=seed, seed2=seed2,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "seed", _op._get_attr_int("seed"),
"seed2", _op._get_attr_int("seed2"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"RandomCrop", _inputs_flat, _attrs, _result)
_result, = _result
return _result
RandomCrop = tf_export("raw_ops.RandomCrop")(_ops.to_raw_op(random_crop))
def random_crop_eager_fallback(image: Annotated[Any, TV_RandomCrop_T], size: Annotated[Any, _atypes.Int64], seed: int, seed2: int, name, ctx) -> Annotated[Any, TV_RandomCrop_T]:
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
_attr_T, (image,) = _execute.args_to_matching_eager([image], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, _dtypes.float32, _dtypes.float64, ])
size = _ops.convert_to_tensor(size, _dtypes.int64)
_inputs_flat = [image, size]
_attrs = ("T", _attr_T, "seed", seed, "seed2", seed2)
_result = _execute.execute(b"RandomCrop", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"RandomCrop", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeArea_T = TypeVar("TV_ResizeArea_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def resize_area(images: Annotated[Any, TV_ResizeArea_T], size: Annotated[Any, _atypes.Int32], align_corners:bool=False, name=None) -> Annotated[Any, _atypes.Float32]:
r"""Resize `images` to `size` using area interpolation.
Input images can be of different types but output images are always float.
The range of pixel values for the output image might be slightly different
from the range for the input image because of limited numerical precision.
To guarantee an output range, for example `[0.0, 1.0]`, apply
`tf.clip_by_value` to the output.
Each output pixel is computed by first transforming the pixel's footprint into
the input tensor and then averaging the pixels that intersect the footprint. An
input pixel's contribution to the average is weighted by the fraction of its
area that intersects the footprint. This is the same as OpenCV's INTER_AREA.
Args:
images: A `Tensor`. Must be one of the following types: `int8`, `uint8`, `int16`, `uint16`, `int32`, `int64`, `half`, `float32`, `float64`, `bfloat16`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and output tensors are
aligned, preserving the values at the corner pixels. Defaults to false.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeArea", name, images, size, "align_corners",
align_corners)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_area_eager_fallback(
images, size, align_corners=align_corners, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeArea", images=images, size=size, align_corners=align_corners,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeArea", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeArea = tf_export("raw_ops.ResizeArea")(_ops.to_raw_op(resize_area))
def resize_area_eager_fallback(images: Annotated[Any, TV_ResizeArea_T], size: Annotated[Any, _atypes.Int32], align_corners: bool, name, ctx) -> Annotated[Any, _atypes.Float32]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.int8, _dtypes.uint8, _dtypes.int16, _dtypes.uint16, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.float32, _dtypes.float64, _dtypes.bfloat16, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
_inputs_flat = [images, size]
_attrs = ("T", _attr_T, "align_corners", align_corners)
_result = _execute.execute(b"ResizeArea", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeArea", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeBicubic_T = TypeVar("TV_ResizeBicubic_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def resize_bicubic(images: Annotated[Any, TV_ResizeBicubic_T], size: Annotated[Any, _atypes.Int32], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, _atypes.Float32]:
r"""Resize `images` to `size` using bicubic interpolation.
Input images can be of different types but output images are always float.
Args:
images: A `Tensor`. Must be one of the following types: `int8`, `uint8`, `int16`, `uint16`, `int32`, `int64`, `half`, `float32`, `float64`, `bfloat16`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and output tensors are
aligned, preserving the values at the corner pixels. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeBicubic", name, images, size, "align_corners",
align_corners, "half_pixel_centers", half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_bicubic_eager_fallback(
images, size, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeBicubic", images=images, size=size,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeBicubic", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeBicubic = tf_export("raw_ops.ResizeBicubic")(_ops.to_raw_op(resize_bicubic))
def resize_bicubic_eager_fallback(images: Annotated[Any, TV_ResizeBicubic_T], size: Annotated[Any, _atypes.Int32], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, _atypes.Float32]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.int8, _dtypes.uint8, _dtypes.int16, _dtypes.uint16, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.float32, _dtypes.float64, _dtypes.bfloat16, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
_inputs_flat = [images, size]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeBicubic", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeBicubic", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeBicubicGrad_T = TypeVar("TV_ResizeBicubicGrad_T", _atypes.Float32, _atypes.Float64)
def resize_bicubic_grad(grads: Annotated[Any, _atypes.Float32], original_image: Annotated[Any, TV_ResizeBicubicGrad_T], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, TV_ResizeBicubicGrad_T]:
r"""Computes the gradient of bicubic interpolation.
Args:
grads: A `Tensor` of type `float32`.
4-D with shape `[batch, height, width, channels]`.
original_image: A `Tensor`. Must be one of the following types: `float32`, `float64`.
4-D with shape `[batch, orig_height, orig_width, channels]`,
The image tensor that was resized.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and grad tensors are
aligned. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `original_image`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeBicubicGrad", name, grads, original_image,
"align_corners", align_corners, "half_pixel_centers",
half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_bicubic_grad_eager_fallback(
grads, original_image, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeBicubicGrad", grads=grads, original_image=original_image,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeBicubicGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeBicubicGrad = tf_export("raw_ops.ResizeBicubicGrad")(_ops.to_raw_op(resize_bicubic_grad))
def resize_bicubic_grad_eager_fallback(grads: Annotated[Any, _atypes.Float32], original_image: Annotated[Any, TV_ResizeBicubicGrad_T], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, TV_ResizeBicubicGrad_T]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (original_image,) = _execute.args_to_matching_eager([original_image], ctx, [_dtypes.float32, _dtypes.float64, ])
grads = _ops.convert_to_tensor(grads, _dtypes.float32)
_inputs_flat = [grads, original_image]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeBicubicGrad", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeBicubicGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeBilinear_T = TypeVar("TV_ResizeBilinear_T", _atypes.BFloat16, _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def resize_bilinear(images: Annotated[Any, TV_ResizeBilinear_T], size: Annotated[Any, _atypes.Int32], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, _atypes.Float32]:
r"""Resize `images` to `size` using bilinear interpolation.
Input images can be of different types but output images are always float.
Args:
images: A `Tensor`. Must be one of the following types: `int8`, `uint8`, `int16`, `uint16`, `int32`, `int64`, `bfloat16`, `half`, `float32`, `float64`, `bfloat16`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and output tensors are
aligned, preserving the values at the corner pixels. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeBilinear", name, images, size, "align_corners",
align_corners, "half_pixel_centers", half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_bilinear_eager_fallback(
images, size, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeBilinear", images=images, size=size,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeBilinear", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeBilinear = tf_export("raw_ops.ResizeBilinear")(_ops.to_raw_op(resize_bilinear))
def resize_bilinear_eager_fallback(images: Annotated[Any, TV_ResizeBilinear_T], size: Annotated[Any, _atypes.Int32], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, _atypes.Float32]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.int8, _dtypes.uint8, _dtypes.int16, _dtypes.uint16, _dtypes.int32, _dtypes.int64, _dtypes.bfloat16, _dtypes.half, _dtypes.float32, _dtypes.float64, _dtypes.bfloat16, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
_inputs_flat = [images, size]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeBilinear", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeBilinear", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeBilinearGrad_T = TypeVar("TV_ResizeBilinearGrad_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half)
def resize_bilinear_grad(grads: Annotated[Any, _atypes.Float32], original_image: Annotated[Any, TV_ResizeBilinearGrad_T], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, TV_ResizeBilinearGrad_T]:
r"""Computes the gradient of bilinear interpolation.
Args:
grads: A `Tensor` of type `float32`.
4-D with shape `[batch, height, width, channels]`.
original_image: A `Tensor`. Must be one of the following types: `float32`, `bfloat16`, `half`, `float64`.
4-D with shape `[batch, orig_height, orig_width, channels]`,
The image tensor that was resized.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and grad tensors are
aligned. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `original_image`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeBilinearGrad", name, grads, original_image,
"align_corners", align_corners, "half_pixel_centers",
half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_bilinear_grad_eager_fallback(
grads, original_image, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeBilinearGrad", grads=grads, original_image=original_image,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeBilinearGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeBilinearGrad = tf_export("raw_ops.ResizeBilinearGrad")(_ops.to_raw_op(resize_bilinear_grad))
def resize_bilinear_grad_eager_fallback(grads: Annotated[Any, _atypes.Float32], original_image: Annotated[Any, TV_ResizeBilinearGrad_T], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, TV_ResizeBilinearGrad_T]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (original_image,) = _execute.args_to_matching_eager([original_image], ctx, [_dtypes.float32, _dtypes.bfloat16, _dtypes.half, _dtypes.float64, ])
grads = _ops.convert_to_tensor(grads, _dtypes.float32)
_inputs_flat = [grads, original_image]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeBilinearGrad", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeBilinearGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeNearestNeighbor_T = TypeVar("TV_ResizeNearestNeighbor_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def resize_nearest_neighbor(images: Annotated[Any, TV_ResizeNearestNeighbor_T], size: Annotated[Any, _atypes.Int32], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, TV_ResizeNearestNeighbor_T]:
r"""Resize `images` to `size` using nearest neighbor interpolation.
Args:
images: A `Tensor`. Must be one of the following types: `int8`, `uint8`, `int16`, `uint16`, `int32`, `int64`, `half`, `float32`, `float64`, `bfloat16`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `new_height, new_width`. The
new size for the images.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and output tensors are
aligned, preserving the values at the corner pixels. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `images`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeNearestNeighbor", name, images, size, "align_corners",
align_corners, "half_pixel_centers", half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_nearest_neighbor_eager_fallback(
images, size, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeNearestNeighbor", images=images, size=size,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeNearestNeighbor", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeNearestNeighbor = tf_export("raw_ops.ResizeNearestNeighbor")(_ops.to_raw_op(resize_nearest_neighbor))
def resize_nearest_neighbor_eager_fallback(images: Annotated[Any, TV_ResizeNearestNeighbor_T], size: Annotated[Any, _atypes.Int32], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, TV_ResizeNearestNeighbor_T]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.int8, _dtypes.uint8, _dtypes.int16, _dtypes.uint16, _dtypes.int32, _dtypes.int64, _dtypes.half, _dtypes.float32, _dtypes.float64, _dtypes.bfloat16, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
_inputs_flat = [images, size]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeNearestNeighbor", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeNearestNeighbor", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ResizeNearestNeighborGrad_T = TypeVar("TV_ResizeNearestNeighborGrad_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int32, _atypes.Int8, _atypes.UInt8)
def resize_nearest_neighbor_grad(grads: Annotated[Any, TV_ResizeNearestNeighborGrad_T], size: Annotated[Any, _atypes.Int32], align_corners:bool=False, half_pixel_centers:bool=False, name=None) -> Annotated[Any, TV_ResizeNearestNeighborGrad_T]:
r"""Computes the gradient of nearest neighbor interpolation.
Args:
grads: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int32`, `half`, `float32`, `float64`, `bfloat16`.
4-D with shape `[batch, height, width, channels]`.
size: A 1-D int32 Tensor of 2 elements: `orig_height, orig_width`. The
original input size.
align_corners: An optional `bool`. Defaults to `False`.
If true, the centers of the 4 corner pixels of the input and grad tensors are
aligned. Defaults to false.
half_pixel_centers: An optional `bool`. Defaults to `False`.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `grads`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ResizeNearestNeighborGrad", name, grads, size, "align_corners",
align_corners, "half_pixel_centers", half_pixel_centers)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return resize_nearest_neighbor_grad_eager_fallback(
grads, size, align_corners=align_corners,
half_pixel_centers=half_pixel_centers, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ResizeNearestNeighborGrad", grads=grads, size=size,
align_corners=align_corners,
half_pixel_centers=half_pixel_centers,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "align_corners",
_op._get_attr_bool("align_corners"), "half_pixel_centers",
_op._get_attr_bool("half_pixel_centers"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ResizeNearestNeighborGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ResizeNearestNeighborGrad = tf_export("raw_ops.ResizeNearestNeighborGrad")(_ops.to_raw_op(resize_nearest_neighbor_grad))
def resize_nearest_neighbor_grad_eager_fallback(grads: Annotated[Any, TV_ResizeNearestNeighborGrad_T], size: Annotated[Any, _atypes.Int32], align_corners: bool, half_pixel_centers: bool, name, ctx) -> Annotated[Any, TV_ResizeNearestNeighborGrad_T]:
if align_corners is None:
align_corners = False
align_corners = _execute.make_bool(align_corners, "align_corners")
if half_pixel_centers is None:
half_pixel_centers = False
half_pixel_centers = _execute.make_bool(half_pixel_centers, "half_pixel_centers")
_attr_T, (grads,) = _execute.args_to_matching_eager([grads], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int32, _dtypes.half, _dtypes.float32, _dtypes.float64, _dtypes.bfloat16, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
_inputs_flat = [grads, size]
_attrs = ("T", _attr_T, "align_corners", align_corners,
"half_pixel_centers", half_pixel_centers)
_result = _execute.execute(b"ResizeNearestNeighborGrad", 1,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ResizeNearestNeighborGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_SampleDistortedBoundingBoxOutput = collections.namedtuple(
"SampleDistortedBoundingBox",
["begin", "size", "bboxes"])
TV_SampleDistortedBoundingBox_T = TypeVar("TV_SampleDistortedBoundingBox_T", _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt8)
def sample_distorted_bounding_box(image_size: Annotated[Any, TV_SampleDistortedBoundingBox_T], bounding_boxes: Annotated[Any, _atypes.Float32], seed:int=0, seed2:int=0, min_object_covered:float=0.1, aspect_ratio_range=[0.75, 1.33], area_range=[0.05, 1], max_attempts:int=100, use_image_if_no_bounding_boxes:bool=False, name=None):
r"""Generate a single randomly distorted bounding box for an image.
Bounding box annotations are often supplied in addition to ground-truth labels
in image recognition or object localization tasks. A common technique for
training such a system is to randomly distort an image while preserving
its content, i.e. *data augmentation*. This Op outputs a randomly distorted
localization of an object, i.e. bounding box, given an `image_size`,
`bounding_boxes` and a series of constraints.
The output of this Op is a single bounding box that may be used to crop the
original image. The output is returned as 3 tensors: `begin`, `size` and
`bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the
image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize
what the bounding box looks like.
Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The
bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and
height of the underlying image.
For example,
```python
# Generate a single distorted bounding box.
begin, size, bbox_for_draw = tf.image.sample_distorted_bounding_box(
tf.shape(image),
bounding_boxes=bounding_boxes)
# Draw the bounding box in an image summary.
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
bbox_for_draw)
tf.summary.image('images_with_box', image_with_box)
# Employ the bounding box to distort the image.
distorted_image = tf.slice(image, begin, size)
```
Note that if no bounding box information is available, setting
`use_image_if_no_bounding_boxes = true` will assume there is a single implicit
bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is
false and no bounding boxes are supplied, an error is raised.
Args:
image_size: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int16`, `int32`, `int64`.
1-D, containing `[height, width, channels]`.
bounding_boxes: A `Tensor` of type `float32`.
3-D with shape `[batch, N, 4]` describing the N bounding boxes
associated with the image.
seed: An optional `int`. Defaults to `0`.
If either `seed` or `seed2` are set to non-zero, the random number
generator is seeded by the given `seed`. Otherwise, it is seeded by a random
seed.
seed2: An optional `int`. Defaults to `0`.
A second seed to avoid seed collision.
min_object_covered: An optional `float`. Defaults to `0.1`.
The cropped area of the image must contain at least this
fraction of any bounding box supplied. The value of this parameter should be
non-negative. In the case of 0, the cropped area does not need to overlap
any of the bounding boxes supplied.
aspect_ratio_range: An optional list of `floats`. Defaults to `[0.75, 1.33]`.
The cropped area of the image must have an aspect ratio =
width / height within this range.
area_range: An optional list of `floats`. Defaults to `[0.05, 1]`.
The cropped area of the image must contain a fraction of the
supplied image within this range.
max_attempts: An optional `int`. Defaults to `100`.
Number of attempts at generating a cropped region of the image
of the specified constraints. After `max_attempts` failures, return the entire
image.
use_image_if_no_bounding_boxes: An optional `bool`. Defaults to `False`.
Controls behavior if no bounding boxes supplied.
If true, assume an implicit bounding box covering the whole input. If false,
raise an error.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (begin, size, bboxes).
begin: A `Tensor`. Has the same type as `image_size`.
size: A `Tensor`. Has the same type as `image_size`.
bboxes: A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "SampleDistortedBoundingBox", name, image_size, bounding_boxes,
"seed", seed, "seed2", seed2, "min_object_covered",
min_object_covered, "aspect_ratio_range", aspect_ratio_range,
"area_range", area_range, "max_attempts", max_attempts,
"use_image_if_no_bounding_boxes", use_image_if_no_bounding_boxes)
_result = _SampleDistortedBoundingBoxOutput._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return sample_distorted_bounding_box_eager_fallback(
image_size, bounding_boxes, seed=seed, seed2=seed2,
min_object_covered=min_object_covered,
aspect_ratio_range=aspect_ratio_range, area_range=area_range,
max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
if min_object_covered is None:
min_object_covered = 0.1
min_object_covered = _execute.make_float(min_object_covered, "min_object_covered")
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'sample_distorted_bounding_box' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'sample_distorted_bounding_box' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"SampleDistortedBoundingBox", image_size=image_size,
bounding_boxes=bounding_boxes,
seed=seed, seed2=seed2,
min_object_covered=min_object_covered,
aspect_ratio_range=aspect_ratio_range,
area_range=area_range,
max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "seed", _op._get_attr_int("seed"),
"seed2", _op._get_attr_int("seed2"), "min_object_covered",
_op.get_attr("min_object_covered"), "aspect_ratio_range",
_op.get_attr("aspect_ratio_range"), "area_range",
_op.get_attr("area_range"), "max_attempts",
_op._get_attr_int("max_attempts"),
"use_image_if_no_bounding_boxes",
_op._get_attr_bool("use_image_if_no_bounding_boxes"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"SampleDistortedBoundingBox", _inputs_flat, _attrs, _result)
_result = _SampleDistortedBoundingBoxOutput._make(_result)
return _result
SampleDistortedBoundingBox = tf_export("raw_ops.SampleDistortedBoundingBox")(_ops.to_raw_op(sample_distorted_bounding_box))
def sample_distorted_bounding_box_eager_fallback(image_size: Annotated[Any, TV_SampleDistortedBoundingBox_T], bounding_boxes: Annotated[Any, _atypes.Float32], seed: int, seed2: int, min_object_covered: float, aspect_ratio_range, area_range, max_attempts: int, use_image_if_no_bounding_boxes: bool, name, ctx):
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
if min_object_covered is None:
min_object_covered = 0.1
min_object_covered = _execute.make_float(min_object_covered, "min_object_covered")
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'sample_distorted_bounding_box' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'sample_distorted_bounding_box' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_attr_T, (image_size,) = _execute.args_to_matching_eager([image_size], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, ])
bounding_boxes = _ops.convert_to_tensor(bounding_boxes, _dtypes.float32)
_inputs_flat = [image_size, bounding_boxes]
_attrs = ("T", _attr_T, "seed", seed, "seed2", seed2, "min_object_covered",
min_object_covered, "aspect_ratio_range", aspect_ratio_range, "area_range",
area_range, "max_attempts", max_attempts, "use_image_if_no_bounding_boxes",
use_image_if_no_bounding_boxes)
_result = _execute.execute(b"SampleDistortedBoundingBox", 3,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"SampleDistortedBoundingBox", _inputs_flat, _attrs, _result)
_result = _SampleDistortedBoundingBoxOutput._make(_result)
return _result
_SampleDistortedBoundingBoxV2Output = collections.namedtuple(
"SampleDistortedBoundingBoxV2",
["begin", "size", "bboxes"])
TV_SampleDistortedBoundingBoxV2_T = TypeVar("TV_SampleDistortedBoundingBoxV2_T", _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt8)
def sample_distorted_bounding_box_v2(image_size: Annotated[Any, TV_SampleDistortedBoundingBoxV2_T], bounding_boxes: Annotated[Any, _atypes.Float32], min_object_covered: Annotated[Any, _atypes.Float32], seed:int=0, seed2:int=0, aspect_ratio_range=[0.75, 1.33], area_range=[0.05, 1], max_attempts:int=100, use_image_if_no_bounding_boxes:bool=False, name=None):
r"""Generate a single randomly distorted bounding box for an image.
Bounding box annotations are often supplied in addition to ground-truth labels
in image recognition or object localization tasks. A common technique for
training such a system is to randomly distort an image while preserving
its content, i.e. *data augmentation*. This Op outputs a randomly distorted
localization of an object, i.e. bounding box, given an `image_size`,
`bounding_boxes` and a series of constraints.
The output of this Op is a single bounding box that may be used to crop the
original image. The output is returned as 3 tensors: `begin`, `size` and
`bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the
image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize
what the bounding box looks like.
Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The
bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and
height of the underlying image.
For example,
```python
# Generate a single distorted bounding box.
begin, size, bbox_for_draw = tf.image.sample_distorted_bounding_box(
tf.shape(image),
bounding_boxes=bounding_boxes)
# Draw the bounding box in an image summary.
image_with_box = tf.image.draw_bounding_boxes(tf.expand_dims(image, 0),
bbox_for_draw)
tf.summary.image('images_with_box', image_with_box)
# Employ the bounding box to distort the image.
distorted_image = tf.slice(image, begin, size)
```
Note that if no bounding box information is available, setting
`use_image_if_no_bounding_boxes = true` will assume there is a single implicit
bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is
false and no bounding boxes are supplied, an error is raised.
Args:
image_size: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int16`, `int32`, `int64`.
1-D, containing `[height, width, channels]`.
bounding_boxes: A `Tensor` of type `float32`.
3-D with shape `[batch, N, 4]` describing the N bounding boxes
associated with the image.
min_object_covered: A `Tensor` of type `float32`.
The cropped area of the image must contain at least this
fraction of any bounding box supplied. The value of this parameter should be
non-negative. In the case of 0, the cropped area does not need to overlap
any of the bounding boxes supplied.
seed: An optional `int`. Defaults to `0`.
If either `seed` or `seed2` are set to non-zero, the random number
generator is seeded by the given `seed`. Otherwise, it is seeded by a random
seed.
seed2: An optional `int`. Defaults to `0`.
A second seed to avoid seed collision.
aspect_ratio_range: An optional list of `floats`. Defaults to `[0.75, 1.33]`.
The cropped area of the image must have an aspect ratio =
width / height within this range.
area_range: An optional list of `floats`. Defaults to `[0.05, 1]`.
The cropped area of the image must contain a fraction of the
supplied image within this range.
max_attempts: An optional `int`. Defaults to `100`.
Number of attempts at generating a cropped region of the image
of the specified constraints. After `max_attempts` failures, return the entire
image.
use_image_if_no_bounding_boxes: An optional `bool`. Defaults to `False`.
Controls behavior if no bounding boxes supplied.
If true, assume an implicit bounding box covering the whole input. If false,
raise an error.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (begin, size, bboxes).
begin: A `Tensor`. Has the same type as `image_size`.
size: A `Tensor`. Has the same type as `image_size`.
bboxes: A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "SampleDistortedBoundingBoxV2", name, image_size,
bounding_boxes, min_object_covered, "seed", seed, "seed2", seed2,
"aspect_ratio_range", aspect_ratio_range, "area_range", area_range,
"max_attempts", max_attempts, "use_image_if_no_bounding_boxes",
use_image_if_no_bounding_boxes)
_result = _SampleDistortedBoundingBoxV2Output._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return sample_distorted_bounding_box_v2_eager_fallback(
image_size, bounding_boxes, min_object_covered, seed=seed,
seed2=seed2, aspect_ratio_range=aspect_ratio_range,
area_range=area_range, max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'sample_distorted_bounding_box_v2' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'sample_distorted_bounding_box_v2' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"SampleDistortedBoundingBoxV2", image_size=image_size,
bounding_boxes=bounding_boxes,
min_object_covered=min_object_covered,
seed=seed, seed2=seed2,
aspect_ratio_range=aspect_ratio_range,
area_range=area_range,
max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "seed", _op._get_attr_int("seed"),
"seed2", _op._get_attr_int("seed2"), "aspect_ratio_range",
_op.get_attr("aspect_ratio_range"), "area_range",
_op.get_attr("area_range"), "max_attempts",
_op._get_attr_int("max_attempts"),
"use_image_if_no_bounding_boxes",
_op._get_attr_bool("use_image_if_no_bounding_boxes"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"SampleDistortedBoundingBoxV2", _inputs_flat, _attrs, _result)
_result = _SampleDistortedBoundingBoxV2Output._make(_result)
return _result
SampleDistortedBoundingBoxV2 = tf_export("raw_ops.SampleDistortedBoundingBoxV2")(_ops.to_raw_op(sample_distorted_bounding_box_v2))
def sample_distorted_bounding_box_v2_eager_fallback(image_size: Annotated[Any, TV_SampleDistortedBoundingBoxV2_T], bounding_boxes: Annotated[Any, _atypes.Float32], min_object_covered: Annotated[Any, _atypes.Float32], seed: int, seed2: int, aspect_ratio_range, area_range, max_attempts: int, use_image_if_no_bounding_boxes: bool, name, ctx):
if seed is None:
seed = 0
seed = _execute.make_int(seed, "seed")
if seed2 is None:
seed2 = 0
seed2 = _execute.make_int(seed2, "seed2")
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'sample_distorted_bounding_box_v2' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'sample_distorted_bounding_box_v2' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_attr_T, (image_size,) = _execute.args_to_matching_eager([image_size], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, ])
bounding_boxes = _ops.convert_to_tensor(bounding_boxes, _dtypes.float32)
min_object_covered = _ops.convert_to_tensor(min_object_covered, _dtypes.float32)
_inputs_flat = [image_size, bounding_boxes, min_object_covered]
_attrs = ("T", _attr_T, "seed", seed, "seed2", seed2, "aspect_ratio_range",
aspect_ratio_range, "area_range", area_range, "max_attempts", max_attempts,
"use_image_if_no_bounding_boxes", use_image_if_no_bounding_boxes)
_result = _execute.execute(b"SampleDistortedBoundingBoxV2", 3,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"SampleDistortedBoundingBoxV2", _inputs_flat, _attrs, _result)
_result = _SampleDistortedBoundingBoxV2Output._make(_result)
return _result
TV_ScaleAndTranslate_T = TypeVar("TV_ScaleAndTranslate_T", _atypes.BFloat16, _atypes.Float32, _atypes.Float64, _atypes.Half, _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt16, _atypes.UInt8)
def scale_and_translate(images: Annotated[Any, TV_ScaleAndTranslate_T], size: Annotated[Any, _atypes.Int32], scale: Annotated[Any, _atypes.Float32], translation: Annotated[Any, _atypes.Float32], kernel_type:str="lanczos3", antialias:bool=True, name=None) -> Annotated[Any, _atypes.Float32]:
r"""TODO: add doc.
Args:
images: A `Tensor`. Must be one of the following types: `int8`, `uint8`, `int16`, `uint16`, `int32`, `int64`, `bfloat16`, `half`, `float32`, `float64`.
size: A `Tensor` of type `int32`.
scale: A `Tensor` of type `float32`.
translation: A `Tensor` of type `float32`.
kernel_type: An optional `string`. Defaults to `"lanczos3"`.
antialias: An optional `bool`. Defaults to `True`.
name: A name for the operation (optional).
Returns:
A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ScaleAndTranslate", name, images, size, scale, translation,
"kernel_type", kernel_type, "antialias", antialias)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return scale_and_translate_eager_fallback(
images, size, scale, translation, kernel_type=kernel_type,
antialias=antialias, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if kernel_type is None:
kernel_type = "lanczos3"
kernel_type = _execute.make_str(kernel_type, "kernel_type")
if antialias is None:
antialias = True
antialias = _execute.make_bool(antialias, "antialias")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ScaleAndTranslate", images=images, size=size, scale=scale,
translation=translation, kernel_type=kernel_type,
antialias=antialias, name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "kernel_type",
_op.get_attr("kernel_type"), "antialias",
_op._get_attr_bool("antialias"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ScaleAndTranslate", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ScaleAndTranslate = tf_export("raw_ops.ScaleAndTranslate")(_ops.to_raw_op(scale_and_translate))
def scale_and_translate_eager_fallback(images: Annotated[Any, TV_ScaleAndTranslate_T], size: Annotated[Any, _atypes.Int32], scale: Annotated[Any, _atypes.Float32], translation: Annotated[Any, _atypes.Float32], kernel_type: str, antialias: bool, name, ctx) -> Annotated[Any, _atypes.Float32]:
if kernel_type is None:
kernel_type = "lanczos3"
kernel_type = _execute.make_str(kernel_type, "kernel_type")
if antialias is None:
antialias = True
antialias = _execute.make_bool(antialias, "antialias")
_attr_T, (images,) = _execute.args_to_matching_eager([images], ctx, [_dtypes.int8, _dtypes.uint8, _dtypes.int16, _dtypes.uint16, _dtypes.int32, _dtypes.int64, _dtypes.bfloat16, _dtypes.half, _dtypes.float32, _dtypes.float64, ])
size = _ops.convert_to_tensor(size, _dtypes.int32)
scale = _ops.convert_to_tensor(scale, _dtypes.float32)
translation = _ops.convert_to_tensor(translation, _dtypes.float32)
_inputs_flat = [images, size, scale, translation]
_attrs = ("T", _attr_T, "kernel_type", kernel_type, "antialias", antialias)
_result = _execute.execute(b"ScaleAndTranslate", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ScaleAndTranslate", _inputs_flat, _attrs, _result)
_result, = _result
return _result
TV_ScaleAndTranslateGrad_T = TypeVar("TV_ScaleAndTranslateGrad_T", bound=_atypes.Float32)
def scale_and_translate_grad(grads: Annotated[Any, TV_ScaleAndTranslateGrad_T], original_image: Annotated[Any, TV_ScaleAndTranslateGrad_T], scale: Annotated[Any, _atypes.Float32], translation: Annotated[Any, _atypes.Float32], kernel_type:str="lanczos3", antialias:bool=True, name=None) -> Annotated[Any, TV_ScaleAndTranslateGrad_T]:
r"""TODO: add doc.
Args:
grads: A `Tensor`. Must be one of the following types: `float32`.
original_image: A `Tensor`. Must have the same type as `grads`.
scale: A `Tensor` of type `float32`.
translation: A `Tensor` of type `float32`.
kernel_type: An optional `string`. Defaults to `"lanczos3"`.
antialias: An optional `bool`. Defaults to `True`.
name: A name for the operation (optional).
Returns:
A `Tensor`. Has the same type as `grads`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "ScaleAndTranslateGrad", name, grads, original_image, scale,
translation, "kernel_type", kernel_type, "antialias", antialias)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return scale_and_translate_grad_eager_fallback(
grads, original_image, scale, translation, kernel_type=kernel_type,
antialias=antialias, name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if kernel_type is None:
kernel_type = "lanczos3"
kernel_type = _execute.make_str(kernel_type, "kernel_type")
if antialias is None:
antialias = True
antialias = _execute.make_bool(antialias, "antialias")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"ScaleAndTranslateGrad", grads=grads, original_image=original_image,
scale=scale, translation=translation,
kernel_type=kernel_type, antialias=antialias,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "kernel_type",
_op.get_attr("kernel_type"), "antialias",
_op._get_attr_bool("antialias"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"ScaleAndTranslateGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
ScaleAndTranslateGrad = tf_export("raw_ops.ScaleAndTranslateGrad")(_ops.to_raw_op(scale_and_translate_grad))
def scale_and_translate_grad_eager_fallback(grads: Annotated[Any, TV_ScaleAndTranslateGrad_T], original_image: Annotated[Any, TV_ScaleAndTranslateGrad_T], scale: Annotated[Any, _atypes.Float32], translation: Annotated[Any, _atypes.Float32], kernel_type: str, antialias: bool, name, ctx) -> Annotated[Any, TV_ScaleAndTranslateGrad_T]:
if kernel_type is None:
kernel_type = "lanczos3"
kernel_type = _execute.make_str(kernel_type, "kernel_type")
if antialias is None:
antialias = True
antialias = _execute.make_bool(antialias, "antialias")
_attr_T, _inputs_T = _execute.args_to_matching_eager([grads, original_image], ctx, [_dtypes.float32, ])
(grads, original_image) = _inputs_T
scale = _ops.convert_to_tensor(scale, _dtypes.float32)
translation = _ops.convert_to_tensor(translation, _dtypes.float32)
_inputs_flat = [grads, original_image, scale, translation]
_attrs = ("T", _attr_T, "kernel_type", kernel_type, "antialias", antialias)
_result = _execute.execute(b"ScaleAndTranslateGrad", 1, inputs=_inputs_flat,
attrs=_attrs, ctx=ctx, name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"ScaleAndTranslateGrad", _inputs_flat, _attrs, _result)
_result, = _result
return _result
_StatelessSampleDistortedBoundingBoxOutput = collections.namedtuple(
"StatelessSampleDistortedBoundingBox",
["begin", "size", "bboxes"])
TV_StatelessSampleDistortedBoundingBox_T = TypeVar("TV_StatelessSampleDistortedBoundingBox_T", _atypes.Int16, _atypes.Int32, _atypes.Int64, _atypes.Int8, _atypes.UInt8)
TV_StatelessSampleDistortedBoundingBox_Tseed = TypeVar("TV_StatelessSampleDistortedBoundingBox_Tseed", _atypes.Int32, _atypes.Int64)
def stateless_sample_distorted_bounding_box(image_size: Annotated[Any, TV_StatelessSampleDistortedBoundingBox_T], bounding_boxes: Annotated[Any, _atypes.Float32], min_object_covered: Annotated[Any, _atypes.Float32], seed: Annotated[Any, TV_StatelessSampleDistortedBoundingBox_Tseed], aspect_ratio_range=[0.75, 1.33], area_range=[0.05, 1], max_attempts:int=100, use_image_if_no_bounding_boxes:bool=False, name=None):
r"""Generate a randomly distorted bounding box for an image deterministically.
Bounding box annotations are often supplied in addition to ground-truth labels
in image recognition or object localization tasks. A common technique for
training such a system is to randomly distort an image while preserving its
content, i.e. *data augmentation*. This Op, given the same `seed`,
deterministically outputs a randomly distorted localization of an object, i.e.
bounding box, given an `image_size`, `bounding_boxes` and a series of
constraints.
The output of this Op is a single bounding box that may be used to crop the
original image. The output is returned as 3 tensors: `begin`, `size` and
`bboxes`. The first 2 tensors can be fed directly into `tf.slice` to crop the
image. The latter may be supplied to `tf.image.draw_bounding_boxes` to visualize
what the bounding box looks like.
Bounding boxes are supplied and returned as `[y_min, x_min, y_max, x_max]`. The
bounding box coordinates are floats in `[0.0, 1.0]` relative to the width and
the height of the underlying image.
The output of this Op is guaranteed to be the same given the same `seed` and is
independent of how many times the function is called, and independent of global
seed settings (e.g. `tf.random.set_seed`).
Example usage:
>>> image = np.array([[[1], [2], [3]], [[4], [5], [6]], [[7], [8], [9]]])
>>> bbox = tf.constant(
... [0.0, 0.0, 1.0, 1.0], dtype=tf.float32, shape=[1, 1, 4])
>>> seed = (1, 2)
>>> # Generate a single distorted bounding box.
>>> bbox_begin, bbox_size, bbox_draw = (
... tf.image.stateless_sample_distorted_bounding_box(
... tf.shape(image), bounding_boxes=bbox, seed=seed))
>>> # Employ the bounding box to distort the image.
>>> tf.slice(image, bbox_begin, bbox_size)
<tf.Tensor: shape=(2, 2, 1), dtype=int64, numpy=
array([[[1],
[2]],
[[4],
[5]]])>
>>> # Draw the bounding box in an image summary.
>>> colors = np.array([[1.0, 0.0, 0.0], [0.0, 0.0, 1.0]])
>>> tf.image.draw_bounding_boxes(
... tf.expand_dims(tf.cast(image, tf.float32),0), bbox_draw, colors)
<tf.Tensor: shape=(1, 3, 3, 1), dtype=float32, numpy=
array([[[[1.],
[1.],
[3.]],
[[1.],
[1.],
[6.]],
[[7.],
[8.],
[9.]]]], dtype=float32)>
Note that if no bounding box information is available, setting
`use_image_if_no_bounding_boxes = true` will assume there is a single implicit
bounding box covering the whole image. If `use_image_if_no_bounding_boxes` is
false and no bounding boxes are supplied, an error is raised.
Args:
image_size: A `Tensor`. Must be one of the following types: `uint8`, `int8`, `int16`, `int32`, `int64`.
1-D, containing `[height, width, channels]`.
bounding_boxes: A `Tensor` of type `float32`.
3-D with shape `[batch, N, 4]` describing the N bounding boxes
associated with the image.
min_object_covered: A `Tensor` of type `float32`.
The cropped area of the image must contain at least this
fraction of any bounding box supplied. The value of this parameter should be
non-negative. In the case of 0, the cropped area does not need to overlap
any of the bounding boxes supplied.
seed: A `Tensor`. Must be one of the following types: `int32`, `int64`.
1-D with shape `[2]`. The seed to the random number generator. Must have dtype
`int32` or `int64`. (When using XLA, only `int32` is allowed.)
aspect_ratio_range: An optional list of `floats`. Defaults to `[0.75, 1.33]`.
The cropped area of the image must have an aspect ratio =
width / height within this range.
area_range: An optional list of `floats`. Defaults to `[0.05, 1]`.
The cropped area of the image must contain a fraction of the
supplied image within this range.
max_attempts: An optional `int`. Defaults to `100`.
Number of attempts at generating a cropped region of the image
of the specified constraints. After `max_attempts` failures, return the entire
image.
use_image_if_no_bounding_boxes: An optional `bool`. Defaults to `False`.
Controls behavior if no bounding boxes supplied.
If true, assume an implicit bounding box covering the whole input. If false,
raise an error.
name: A name for the operation (optional).
Returns:
A tuple of `Tensor` objects (begin, size, bboxes).
begin: A `Tensor`. Has the same type as `image_size`.
size: A `Tensor`. Has the same type as `image_size`.
bboxes: A `Tensor` of type `float32`.
"""
_ctx = _context._context or _context.context()
tld = _ctx._thread_local_data
if tld.is_eager:
try:
_result = pywrap_tfe.TFE_Py_FastPathExecute(
_ctx, "StatelessSampleDistortedBoundingBox", name, image_size,
bounding_boxes, min_object_covered, seed, "aspect_ratio_range",
aspect_ratio_range, "area_range", area_range, "max_attempts",
max_attempts, "use_image_if_no_bounding_boxes",
use_image_if_no_bounding_boxes)
_result = _StatelessSampleDistortedBoundingBoxOutput._make(_result)
return _result
except _core._NotOkStatusException as e:
_ops.raise_from_not_ok_status(e, name)
except _core._FallbackException:
pass
try:
return stateless_sample_distorted_bounding_box_eager_fallback(
image_size, bounding_boxes, min_object_covered, seed,
aspect_ratio_range=aspect_ratio_range, area_range=area_range,
max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name, ctx=_ctx)
except _core._SymbolicException:
pass # Add nodes to the TensorFlow graph.
# Add nodes to the TensorFlow graph.
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'stateless_sample_distorted_bounding_box' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'stateless_sample_distorted_bounding_box' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_, _, _op, _outputs = _op_def_library._apply_op_helper(
"StatelessSampleDistortedBoundingBox", image_size=image_size,
bounding_boxes=bounding_boxes,
min_object_covered=min_object_covered,
seed=seed,
aspect_ratio_range=aspect_ratio_range,
area_range=area_range,
max_attempts=max_attempts,
use_image_if_no_bounding_boxes=use_image_if_no_bounding_boxes,
name=name)
_result = _outputs[:]
if _execute.must_record_gradient():
_attrs = ("T", _op._get_attr_type("T"), "Tseed",
_op._get_attr_type("Tseed"), "aspect_ratio_range",
_op.get_attr("aspect_ratio_range"), "area_range",
_op.get_attr("area_range"), "max_attempts",
_op._get_attr_int("max_attempts"),
"use_image_if_no_bounding_boxes",
_op._get_attr_bool("use_image_if_no_bounding_boxes"))
_inputs_flat = _op.inputs
_execute.record_gradient(
"StatelessSampleDistortedBoundingBox", _inputs_flat, _attrs, _result)
_result = _StatelessSampleDistortedBoundingBoxOutput._make(_result)
return _result
StatelessSampleDistortedBoundingBox = tf_export("raw_ops.StatelessSampleDistortedBoundingBox")(_ops.to_raw_op(stateless_sample_distorted_bounding_box))
def stateless_sample_distorted_bounding_box_eager_fallback(image_size: Annotated[Any, TV_StatelessSampleDistortedBoundingBox_T], bounding_boxes: Annotated[Any, _atypes.Float32], min_object_covered: Annotated[Any, _atypes.Float32], seed: Annotated[Any, TV_StatelessSampleDistortedBoundingBox_Tseed], aspect_ratio_range, area_range, max_attempts: int, use_image_if_no_bounding_boxes: bool, name, ctx):
if aspect_ratio_range is None:
aspect_ratio_range = [0.75, 1.33]
if not isinstance(aspect_ratio_range, (list, tuple)):
raise TypeError(
"Expected list for 'aspect_ratio_range' argument to "
"'stateless_sample_distorted_bounding_box' Op, not %r." % aspect_ratio_range)
aspect_ratio_range = [_execute.make_float(_f, "aspect_ratio_range") for _f in aspect_ratio_range]
if area_range is None:
area_range = [0.05, 1]
if not isinstance(area_range, (list, tuple)):
raise TypeError(
"Expected list for 'area_range' argument to "
"'stateless_sample_distorted_bounding_box' Op, not %r." % area_range)
area_range = [_execute.make_float(_f, "area_range") for _f in area_range]
if max_attempts is None:
max_attempts = 100
max_attempts = _execute.make_int(max_attempts, "max_attempts")
if use_image_if_no_bounding_boxes is None:
use_image_if_no_bounding_boxes = False
use_image_if_no_bounding_boxes = _execute.make_bool(use_image_if_no_bounding_boxes, "use_image_if_no_bounding_boxes")
_attr_T, (image_size,) = _execute.args_to_matching_eager([image_size], ctx, [_dtypes.uint8, _dtypes.int8, _dtypes.int16, _dtypes.int32, _dtypes.int64, ])
_attr_Tseed, (seed,) = _execute.args_to_matching_eager([seed], ctx, [_dtypes.int32, _dtypes.int64, ])
bounding_boxes = _ops.convert_to_tensor(bounding_boxes, _dtypes.float32)
min_object_covered = _ops.convert_to_tensor(min_object_covered, _dtypes.float32)
_inputs_flat = [image_size, bounding_boxes, min_object_covered, seed]
_attrs = ("T", _attr_T, "Tseed", _attr_Tseed, "aspect_ratio_range",
aspect_ratio_range, "area_range", area_range, "max_attempts", max_attempts,
"use_image_if_no_bounding_boxes", use_image_if_no_bounding_boxes)
_result = _execute.execute(b"StatelessSampleDistortedBoundingBox", 3,
inputs=_inputs_flat, attrs=_attrs, ctx=ctx,
name=name)
if _execute.must_record_gradient():
_execute.record_gradient(
"StatelessSampleDistortedBoundingBox", _inputs_flat, _attrs, _result)
_result = _StatelessSampleDistortedBoundingBoxOutput._make(_result)
return _result