1775 lines
4.5 MiB
Plaintext
1775 lines
4.5 MiB
Plaintext
|
{
|
|||
|
|
"cells": [
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "g_nWetWWd_ns"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"##### Copyright 2018 The TensorFlow Authors."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 1,
|
|||
|
|
"metadata": {
|
|||
|
|
"cellView": "form",
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:40.659665Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:40.659217Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:40.663044Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:40.662455Z"
|
|||
|
|
},
|
|||
|
|
"id": "2pHVBk_seED1"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n",
|
|||
|
|
"# you may not use this file except in compliance with the License.\n",
|
|||
|
|
"# You may obtain a copy of the License at\n",
|
|||
|
|
"#\n",
|
|||
|
|
"# https://www.apache.org/licenses/LICENSE-2.0\n",
|
|||
|
|
"#\n",
|
|||
|
|
"# Unless required by applicable law or agreed to in writing, software\n",
|
|||
|
|
"# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
|
|||
|
|
"# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
|
|||
|
|
"# See the License for the specific language governing permissions and\n",
|
|||
|
|
"# limitations under the License."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "6msVLevwcRhm"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"# Neural style transfer"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "Ds4o1h4WHz9U"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"<table class=\"tfo-notebook-buttons\" align=\"left\">\n",
|
|||
|
|
" <td>\n",
|
|||
|
|
" <a target=\"_blank\" href=\"https://www.tensorflow.org/tutorials/generative/style_transfer\"><img src=\"https://www.tensorflow.org/images/tf_logo_32px.png\" />View on TensorFlow.org</a>\n",
|
|||
|
|
" </td>\n",
|
|||
|
|
" <td>\n",
|
|||
|
|
" <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/generative/style_transfer.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n",
|
|||
|
|
" </td>\n",
|
|||
|
|
" <td>\n",
|
|||
|
|
" <a target=\"_blank\" href=\"https://github.com/tensorflow/docs/blob/master/site/en/tutorials/generative/style_transfer.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View on GitHub</a>\n",
|
|||
|
|
" </td>\n",
|
|||
|
|
" <td>\n",
|
|||
|
|
" <a href=\"https://storage.googleapis.com/tensorflow_docs/docs/site/en/tutorials/generative/style_transfer.ipynb\"><img src=\"https://www.tensorflow.org/images/download_logo_32px.png\" />Download notebook</a>\n",
|
|||
|
|
" </td>\n",
|
|||
|
|
" <td>\n",
|
|||
|
|
" <a href=\"https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2\"><img src=\"https://www.tensorflow.org/images/hub_logo_32px.png\" />See TF Hub model</a>\n",
|
|||
|
|
" </td>\n",
|
|||
|
|
"</table>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "aDyGj8DmXCJI"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"This tutorial uses deep learning to compose one image in the style of another image (ever wish you could paint like Picasso or Van Gogh?). This is known as *neural style transfer* and the technique is outlined in <a href=\"https://arxiv.org/abs/1508.06576\" class=\"external\">A Neural Algorithm of Artistic Style</a> (Gatys et al.). \n",
|
|||
|
|
"\n",
|
|||
|
|
"Note: This tutorial demonstrates the original style-transfer algorithm. It optimizes the image content to a particular style. Modern approaches train a model to generate the stylized image directly (similar to [CycleGAN](./cyclegan.ipynb)). This approach is much faster (up to 1000x).\n",
|
|||
|
|
"\n",
|
|||
|
|
"For a simple application of style transfer with a pretrained model from [TensorFlow Hub](https://tfhub.dev), check out the [Fast style transfer for arbitrary styles](https://www.tensorflow.org/hub/tutorials/tf2_arbitrary_image_stylization) tutorial that uses an [arbitrary image stylization model](https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2). For an example of style transfer with [TensorFlow Lite](https://www.tensorflow.org/lite), refer to [Artistic style transfer with TensorFlow Lite](https://www.tensorflow.org/lite/examples/style_transfer/overview)."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "1b3XwN9V1nvR"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Neural style transfer is an optimization technique used to take two images—a *content* image and a *style reference* image (such as an artwork by a famous painter)—and blend them together so the output image looks like the content image, but “painted” in the style of the style reference image.\n",
|
|||
|
|
"\n",
|
|||
|
|
"This is implemented by optimizing the output image to match the content statistics of the content image and the style statistics of the style reference image. These statistics are extracted from the images using a convolutional network."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "28W8ggyO1KER"
|
|||
|
|
},
|
|||
|
|
"source": []
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "3kb_UJY-jCEl"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"For example, let’s take an image of this dog and Wassily Kandinsky's Composition 7:\n",
|
|||
|
|
"\n",
|
|||
|
|
"<img src=\"https://storage.googleapis.com/download.tensorflow.org/example_images/YellowLabradorLooking_new.jpg\" width=\"500px\"/>\n",
|
|||
|
|
"\n",
|
|||
|
|
"[Yellow Labrador Looking](https://commons.wikimedia.org/wiki/File:YellowLabradorLooking_new.jpg), from Wikimedia Commons by [Elf](https://en.wikipedia.org/wiki/User:Elf). License [CC BY-SA 3.0](https://creativecommons.org/licenses/by-sa/3.0/deed.en)\n",
|
|||
|
|
"\n",
|
|||
|
|
"<img src=\"https://storage.googleapis.com/download.tensorflow.org/example_images/Vassily_Kandinsky%2C_1913_-_Composition_7.jpg\" width=\"500px\"/>\n",
|
|||
|
|
"\n",
|
|||
|
|
"\n",
|
|||
|
|
"Now, what would it look like if Kandinsky decided to paint the picture of this Dog exclusively with this style? Something like this?\n",
|
|||
|
|
"\n",
|
|||
|
|
"<img src=\"https://tensorflow.org/tutorials/generative/images/stylized-image.png\" style=\"width: 500px;\"/>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "U8ajP_u73s6m"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Setup\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "eqxUicSPUOP6"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"### Import and configure modules"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 1,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:40.667189Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:40.666636Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.010101Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.009437Z"
|
|||
|
|
},
|
|||
|
|
"id": "NyftRTSMuwue"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stderr",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"2024-02-23 18:42:03.927415: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
|
|||
|
|
"2024-02-23 18:42:03.955182: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
|
|||
|
|
"2024-02-23 18:42:03.955208: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
|
|||
|
|
"2024-02-23 18:42:03.955840: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
|
|||
|
|
"2024-02-23 18:42:03.960534: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
|||
|
|
"To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
|
|||
|
|
"2024-02-23 18:42:04.492637: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"import os\n",
|
|||
|
|
"import tensorflow as tf\n",
|
|||
|
|
"# Load compressed models from tensorflow_hub\n",
|
|||
|
|
"os.environ['TFHUB_MODEL_LOAD_FORMAT'] = 'COMPRESSED'"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 2,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.014828Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.014014Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.272013Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.271331Z"
|
|||
|
|
},
|
|||
|
|
"id": "sc1OLbOWhPCO"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"import IPython.display as display\n",
|
|||
|
|
"\n",
|
|||
|
|
"import matplotlib.pyplot as plt\n",
|
|||
|
|
"import matplotlib as mpl\n",
|
|||
|
|
"mpl.rcParams['figure.figsize'] = (12, 12)\n",
|
|||
|
|
"mpl.rcParams['axes.grid'] = False\n",
|
|||
|
|
"\n",
|
|||
|
|
"import numpy as np\n",
|
|||
|
|
"import PIL.Image\n",
|
|||
|
|
"import time\n",
|
|||
|
|
"import functools"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 3,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.276044Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.275804Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.279946Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.279369Z"
|
|||
|
|
},
|
|||
|
|
"id": "GM6VEGrGLh62"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def tensor_to_image(tensor):\n",
|
|||
|
|
" tensor = tensor*255\n",
|
|||
|
|
" tensor = np.array(tensor, dtype=np.uint8)\n",
|
|||
|
|
" if np.ndim(tensor)>3:\n",
|
|||
|
|
" assert tensor.shape[0] == 1\n",
|
|||
|
|
" tensor = tensor[0]\n",
|
|||
|
|
" return PIL.Image.fromarray(tensor)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "oeXebYusyHwC"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Download images and choose a style image and a content image:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 4,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.283405Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.283015Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.428345Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.427755Z"
|
|||
|
|
},
|
|||
|
|
"id": "wqc0OJHwyFAk"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"content_path = tf.keras.utils.get_file('YellowLabradorLooking_new.jpg', 'https://storage.googleapis.com/download.tensorflow.org/example_images/YellowLabradorLooking_new.jpg')\n",
|
|||
|
|
"style_path = tf.keras.utils.get_file('kandinsky5.jpg','https://storage.googleapis.com/download.tensorflow.org/example_images/Vassily_Kandinsky%2C_1913_-_Composition_7.jpg')"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "xE4Yt8nArTeR"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Visualize the input"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "klh6ObK2t_vH"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Define a function to load an image and limit its maximum dimension to 512 pixels."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 5,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.431733Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.431458Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.436393Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.435811Z"
|
|||
|
|
},
|
|||
|
|
"id": "3TLljcwv5qZs"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def load_img(path_to_img):\n",
|
|||
|
|
" max_dim = 512\n",
|
|||
|
|
" img = tf.io.read_file(path_to_img)\n",
|
|||
|
|
" img = tf.image.decode_image(img, channels=3)\n",
|
|||
|
|
" img = tf.image.convert_image_dtype(img, tf.float32)\n",
|
|||
|
|
"\n",
|
|||
|
|
" shape = tf.cast(tf.shape(img)[:-1], tf.float32)\n",
|
|||
|
|
" long_dim = max(shape)\n",
|
|||
|
|
" scale = max_dim / long_dim\n",
|
|||
|
|
"\n",
|
|||
|
|
" new_shape = tf.cast(shape * scale, tf.int32)\n",
|
|||
|
|
"\n",
|
|||
|
|
" img = tf.image.resize(img, new_shape)\n",
|
|||
|
|
" img = img[tf.newaxis, :]\n",
|
|||
|
|
" return img"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "2yAlRzJZrWM3"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Create a simple function to display an image:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 6,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.439757Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.439222Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:43.443101Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:43.442492Z"
|
|||
|
|
},
|
|||
|
|
"id": "cBX-eNT8PAK_"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def imshow(image, title=None):\n",
|
|||
|
|
" if len(image.shape) > 3:\n",
|
|||
|
|
" image = tf.squeeze(image, axis=0)\n",
|
|||
|
|
"\n",
|
|||
|
|
" plt.imshow(image)\n",
|
|||
|
|
" if title:\n",
|
|||
|
|
" plt.title(title)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 7,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"chel_path = \"/home/aneta/.ipython/profile_default/neural_style_pytorch/images/chelm.jpg\""
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 8,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"owce_path = \"/home/aneta/.ipython/profile_default/neural_style_pytorch/images/owce.jpg\""
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 9,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:43.446483Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:43.445917Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:46.265458Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:46.264530Z"
|
|||
|
|
},
|
|||
|
|
"id": "_UWQmeEaiKkP"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stderr",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"2024-02-23 18:42:17.937760: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:17.958243: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:17.958276: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:17.962148: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:17.962188: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:17.962200: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:18.072362: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:18.072405: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:18.072411: I tensorflow/core/common_runtime/gpu/gpu_device.cc:2022] Could not identify NUMA node of platform GPU id 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:18.072435: I external/local_xla/xla/stream_executor/cuda/cuda_executor.cc:887] could not open file to read NUMA node: /sys/bus/pci/devices/0000:01:00.0/numa_node\n",
|
|||
|
|
"Your kernel may have been built without NUMA support.\n",
|
|||
|
|
"2024-02-23 18:42:18.072452: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1929] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 1768 MB memory: -> device: 0, name: NVIDIA GeForce RTX 3050 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<Figure size 1200x1200 with 2 Axes>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"content_image = load_img(owce_path)\n",
|
|||
|
|
"style_image = load_img(chel_path)\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.subplot(1, 2, 1)\n",
|
|||
|
|
"imshow(content_image, 'Content Image')\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.subplot(1, 2, 2)\n",
|
|||
|
|
"imshow(style_image, 'Style Image')"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "YMzChXSlKTA2"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Fast Style Transfer using TF-Hub\n",
|
|||
|
|
"\n",
|
|||
|
|
"This tutorial demonstrates the original style-transfer algorithm, which optimizes the image content to a particular style. Before getting into the details, let's see how the [TensorFlow Hub model](https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2) does this:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 43,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:46.276365Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:46.275669Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:34:55.327882Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:34:55.327200Z"
|
|||
|
|
},
|
|||
|
|
"id": "iYSLexgRKSh-"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/jpeg": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAEgAgADASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgcICQoL/8QAtRAAAgEDAwIEAwUFBAQAAAF9AQIDAAQRBRIhMUEGE1FhByJxFDKBkaEII0KxwRVS0fAkM2JyggkKFhcYGRolJicoKSo0NTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqDhIWGh4iJipKTlJWWl5iZmqKjpKWmp6ipqrKztLW2t7i5usLDxMXGx8jJytLT1NXW19jZ2uHi4+Tl5ufo6erx8vP09fb3+Pn6/8QAHwEAAwEBAQEBAQEBAQAAAAAAAAECAwQFBgcICQoL/8QAtREAAgECBAQDBAcFBAQAAQJ3AAECAxEEBSExBhJBUQdhcRMiMoEIFEKRobHBCSMzUvAVYnLRChYkNOEl8RcYGRomJygpKjU2Nzg5OkNERUZHSElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6goOEhYaHiImKkpOUlZaXmJmaoqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4uPk5ebn6Onq8vP09fb3+Pn6/9oADAMBAAIRAxEAPwDp9JLNHJ2wx/nWiQcg+2KzdFOY5SeRvOK1mxxxmuS5221IjjpjpTTknIFPxjpxRj1z+dVcoi8sHqOaTyh/dFSHA9qUYpCsReVgDCiniMHqo4pxxijgd6AsNCgZ+Uc0hjGBhRUqgUfLQBGIu5Ao8oDoo5p5xjrSg+poCwzZgDCigxgjpTyFpORQKw0xg9FFHlkDoKeD1yaMikOwwqvTbRsJwcUuTjk04EZ60wsM2jnK0eXxwMUv0pS3vRoOw0xgAcUgX5s4p27Peg4/+vRoA0KADgdaQpwOKeMUpxignUZ5Y5yKBGB2p30pc89aV0OwwooGNopDGpAwop4BzyaXAAouFmRiMZBPYYoCqBjFPI9KaQT0o0DlAoD2pAmTmnZOKXvRcdn1GEKoxikKAgcU/GetIOM5p3QrdhuzjOKUop6rTiM4waTBpXQJDdvPIoxzwKf1xmlAFFwsRgA9aNgFP24pTRcLDNnAoKqccU4jPSjFCa6By9hgXJzR5YwBipNozQBTug5WMCKTnFN8tQMYqUrxxRtzSuhcpGVHZaNoOOOlSqvY0hTHTvTuFnsR+Wozx1pCi/3RUwXnmjyxSuFmQGIDnFBiDdFFWNlG3FO4WZV+zj0FJ5O0ZxVogAUxwDTuHK2QlVwBgcUhApxGKOKV0g5Ru0YFIQDmncU0g4GKd0OzE2HNGwDginc0q89aLhZkJUHtTTCDjge9WNoHWlKincTTfUqGBeeKEtx5iHAqcpQAQ6/WqulsZuPmVNIZgsoXghjn861w+4c8YrJ0gDM3PBbp+NapTGK50zo5bMU4pOT3pcUYqrhYTHrQBTsZFGKVwsNxS4Bp23NLtouFhnFJzUm2jbjtRcLDMGinFc0oWi4WGj3pTg0u2nBKLgRkdKAtSbaNlADNuaaV9qmCmk2E0ARYoCGpdhpdtAEJXFG3NTbSaXZQBABilxmpvLBpNlAiPaKQJU2zilCigCLaPSk25qbZRsGaAIdppNuDVggUmygZAU9KNlT4FLtBFAFfZSbKsFRQQOKAIAuKNhNT7aMDNAXINppQtTEUmKBEYWlKcVIcCgYoAjEdBjqUkYpQBQBCEpTGKecDpRTAYEpdlPHWigBnl96UIKfSDNFxDNmO1JxUpBpmzJouMacAU0j0qTyzTljHegCuQTTNpq4Yxik8ugZUMeRTRH61e8sU0xD0oAqeVThEKnKECmhTmgCExAUeV6VMQaUKaLgV/LzRs4qwF5oK0xFYoMU1lAx9ateXTWi46U02IoaYhSEkjks38608ZAqrpSho5PZzV9lArLyNCHZ7UpX2qTimtTENCil20UoNABgUuM03mnUCDAo20c85o5oACopAKdgmjHpQAFRigUc0uKAEIFLxSYNLigBKdkDHvTdtAB/KncBcg9KXApNntQOtK4Bxmg8Yo20YouAgNKcUu30puDRcA+lIDin4xRtyaYCFulIM0pWl2igBpOBRnn2pdppdooAjwaXNPwKCvFAEY5ox7U/Ao60AIKTaak2jFAApANxTdvNPIOKBmgBhU4oAx2p5pQCaAGYzRggU/BpaAI9uaCtOzTsUCIwKUAU8LRigBoWgKKeKCPagAwMU0gCnAZpCppgHHFNBp200oUUANPSlGKXbml20ANHWkOcU/bg9KUgUAMC560eWKeMUHpxQBEUFJipDkikx60wGBaXaKUEZpc+lMBuB+dIwOOKeOetDYA56UxGdpZYQMf7zn+daBIbrVLTk/wBHIHBDH+dWyCKyLEI9KTFPxTStABgUmKTmlxmgBcClxTQKcelAheKTik5oOKBjhSimg80ooAUgYpQaTj0pKYhTSbvSjmjHrQA4e9JxnilwKULigBM8UYoIoFABx6UhowaTn0oAUGlxSCg5oAWkpcGnbaAGEnigZp2BS7RQAw0ZzTivrTcUwGnApS1GKXaO9ADRmjpT8CkxSEGelKOOtGBScdjQA7rQM0AUbaBiECgGlKmkxQIU0CjHFLjFACYHcUvFIc+tKAaAEB5pfrRtpcUAIMUvOBQFoxTAUDFBxS4pMe1ABQAc0uR3pufQ0ADYFLkHvTSWxRzQA7NNORSbsdaCy+tACigk46VFv9DS7yaAHlxjrTSw7GmnJoC+tFwAZzzSigD0owR2poB3AobBHtTMnHvQ5ITNUhEOnr+5JPQscVbZRgYqGwUfZgAM4Y1YIrOxdyI5FJnJ61IRTNuKBXGkUgIpxGaaQaAuO69KTNA96MUAIaQ5OKdikOAMmgYgBFKGx3ozmkxzQIkHNHGaQHFIM0CHnFNOexopRigABxTt2elGKQACgAyaUUEjA5pAKYCmkzS0YoAaCadx3pvfg0/NACjFLTeKcDQA3Boz2p+RTcUANJpRmgigECgAwKMA0hpwIHagBuKUAUjOBSCRfWiwDsGjAFMMg9ab5jexoAkLAClByeuahLE0c5pgTkjHWkyKipRxSsFyQkAdaA3pUec0DiiwrkvHegketRFjRnNFguShqXNRZxRuHrRYLkuaNy5qEvTc07DJvMXPWgSA1CRQCB1NAaEzNx1FNDLnrULNnpTdxFILE5kU8A0eYD3qvyaQdaLjsTFyT60wsaQUtK4xVGaeuRTMgU6nYQ7IFLuB6VFk04fSmA8Hnk0dehpnNPFMQuPWmy8JwaXJ7UyYkR5prclhZMBBtU9GOasFh3NczbahcpHjzQAxPbPepjqFy2MSD8qy9ojT2bubu5exzSE5rGXUbgH5nQ/WrMepRsMO+D7CjnTHyNF/miqpvoQoYzDBqRJ45PuyKfxp3RGpJSFsUgdWPDL+dNLr/eX8xVaCH7s0wkHvUL3MSdZVz6A0xrmELuMg/CjQLss59DSbzVUX9uvWT9KrSaxCDiJS59xgUXQa9TXBFLuArE/tmQY+SIDv1p8etxEkSkD02rRoGpsb6XrWS2t2o27Sxz7VF/wkMC/xA+2w5o0Fdm7Sd6pw6nazLkTKPY8VOLqA4xKn50DJcUv4VX+2Q/8APaP/AL6FSrPG3SRD+NFguOyaXdRlT0IP0pDgUAHAqJnPY0jTICR5iZ9NwzUck0SLl5EUepYU7Bck8xqQSv61VW9t2JxMmB70g1G1LYEy07ITbL4c04yN61npqERBxMpwasrcRHH71P8AvoUrBdkxdjjnFIGbPWkDIejqfoRTHniU4MiD6sKARMG96C3oaoNqFsrlTOhx6EGg6jbA8yj8KRRcyT3oxWedYsx92Rj/AMBNQya1GyYjHOfvMpouBq
|
|||
|
|
"image/png": "iVBORw0KGgoAAAANSUhEUgAAAgAAAAEgCAIAAABNXlwGAAEAAElEQVR4AYT963IlSZJmiwWuEZlV1d2H/EWK8IH5AHwSvhMpZzjsmarKyowI3LjWUjXfyJ4jQgfgbqaXTz9VMzf37XsDuPt//j/+7//7//7//r/9X/6vv/7br1+en//yt//zz59vT3f3918+7u7e7u7ev//44/Xj9dsvv95/ufv5/fuXt/enh69f7u7v7h7f394/3j6+fGD55f7p4cuXL29vL/d3D68vL3d3dw9PT29vb6+vb6/vP+/vwPv4eLv7eP/y5ePLl/t73N4+Xt7fX+/ucX+4+/LA/u3tA4eH+y/vdx9PT4+If779fHjE+A2Ax4dHSL2/fXl/fX+4fwTv7suXu4eHcN6/4PL4+Hz/9cs7Bh8fr8Dev79/PJDK3d3r28vHl3egaL/fvd8/3L2+vr6/vT49fjWBL/fvbx/vMnz78uX96fn5y8M9gjfoffl4fni4vyPvd4h/of3wYApYvr+TprRxRmXFYATF+/fXN0ry8Ag33N4+wP7y8e0X6nb38+U7ENTv/p4aPrxRq3eC0HuGxs/Xn5CgPI/3VPSJVMB+fX95eH6iGnQI8OXj3ZoRkj1NS0ph37+YHGEJbFwMx/7+HsLEeJDVPTV5u3v4eHh8hPYje0rz4+fT0/P909P93eOPH39QDghR8+ev3x4fnl9eflJmiiEIKI+PL6+vP99e7xiOLx8gUDa1FJmxu3u0ju8fsr37oICmA8rjExWwoC8/TZ8IDh0zBAPYU88vj3dPsDc41XYMmUXkQjbOhNeX1z++//71l2+Pz4+vry8W4e7DjAG6c+4x4uwf3u8YOaaKzF9fIM3AYcGYvuOD1duLZo8OAbF+/vyDCLp/fNw/Pr1jzPRkZr28P37Ak+I9QPHH+8sX7d/g/vhEeV/BJTb1JCpldw68f1BA9NSTUgt7x1xy4EgTG/wbFir3SPY/X146jwj4/vz0hDHnC7PFfO7pvGJ8D+SXZ4CZIY3n65e7N6rp5OuM+uIgWjGYO3DV6wvTjcn5zkmF7sFTJmuUT5yV7y98Qx7OcHt6/soc+PH9B+f1x5cHQR7vX95/vL39uHv8cv/ITIUMaB/3UPPUAfPpx8+f78y0R0+yu6eHdyYpNh+cLM8PXx5ff74+MGwPDy+vTB6WEWMR+/GJWO8/f7hK4MkJAOmXlxfnALX65kD/+PkHk9lhZ1ghAyqngWOLOXUgVzgywzhdKJYnj98frhCMkoUiYIuSp4+T0A41wlUfR4JRY6Bf3xnmu/fnr8/k4okD4Mt3sv329S/v74/fvwP0qA/ngsP4hYnHifrwyIhAgjOJBeudrEzl4/7dk10NxKFv0Y3PiJHFo6sfQR4fWU4pF4SwZDhcHzwJSIopSB7M4Ps7VpQ3knJhdGrResCd5RFAJIRl2lF6z7bXLy9vloKT4o1pTSUwYDjIh5n++sbZROPLE2cB8/r17fuPn/dPz7/++y+/3D1zDrNZXac+Pta3RYaWSx7R2aNkupu6y6kLynpZTDNh08sqov/C1Ge4MLV4d5zaXlAYBsKUt8YfnASccxJ+eCAsc8KR5vTiAvDlnZnnwFJNXD0RXQtaEEj7nVk4ePhSbcl3zXi6e6Jw8AEBY4ebGUsVmRKcEM4fYyN5vDcoc9Qozu33BxcKje+fHpkllJSLB5k8zTnA6FJWJqGzy1Scjs5KzgcJQJU+A8/2scuC5xOD53z9YAg5XfAjF8cyFpzqLFVPztAvj5WRhYCRIg4VcXXBgYGxtAR0atQgFeeJBF1TgKMC5kGTWHTzxHvXSOsHK+zlgLPogbIUhYMcjQ4yBIyug0fGCIxBxy4cnI1omf7kNYV58Nq4UTodGVCMceNEbkmzPmb+9PgMEH7yJwTnMtDMXOZJjEiAwORuQeMBFUgyZJO9MMRwSnHKoPAsZaHHy6lWNLNx7lhHEgfT88yxttaNF2nCzDPaiwFTlbTwIFvXCSrLTMydxUAZJzJFFoEf4FE2KO4YXGc9XpyWXV6oBmMB+hAxU7CtCNFQCT0gzg8nsOCpZEsIaZmy10MjwmoSgglsnAqU2A12QDEHnUKoZM1mSNYLKkxbnUNp3KiIT5guOdzwODGsTBOHlfj+4UnYOfumTKIyZNxGsZwJBPy7lwXa8vQMk40DsuPFRPGS5nyTUtkzVbChPgjhTIgYs+ib6xP3dpywXQBIz+nsQoqJlWRMlUiTirAOOFQsgybFmoADMgwbKFdkLgBOFkOYLMHiAET2GHJNe3hh1XCGc5dpZAbw/u3hq/PR209QXYmLVfyZPiAw4HLOF0gR2Gq5IhGQKVPuMnfWeI2lzFaezbsTJ4QnOV5iUULStFJOJ+gSmaDRN0nOCo0IQ5YzqqylrFYWBg1BqjM772y86SVrIrLWk8Lj3ePzl8cvj9zxPf/CSgfBUoqKlw0uasJQU4feDWeuOZQSyxZrFDbBkxDQ2g8jWdCBG9Rm4N44/biVIGHMzEryNMmNRZyB9EblWrQZ/eZJZYUJIZgXxjUOICTlSQYtcAjDiDMRLO0XltLWDEouMXCYZLS4HePolAFH3mOtt2tT7Hm1o6ncpPjKtPB6USTTpM2rBMrATYHc4AARbJE7Az+4CXLIu+/hlHN4ue9j5kIDK17reEtLcKeB1wVATOdp1i+lZPHorUibl30BG01ZwQFnR0NfVz5PsnPHDV/zcmTMVIqm55k/X1xp7nxhphtUObp6Wwpfvd1x90+a6J6d8M0gpqoVM0sQzRNfpx8Dai29gONsMAfdgnv2mHkbfBxiEPw2WcrpqHCJb3rkbg1l2qxhCCzQ29u9izlJsnn7IwL0mMsgMCZcYllSLDuFB5ALAJPUF0YMCyuSjKHkBVfSGkLLUxFfx9KAfEGXAWg+WWFKgFHR9qQzee6aYug4QsKpj6xzsDnFiPBF/czdJheqWfl91dQXZdJi4gIGFAtOU1oiEJcyFQYfCCbCjC+Faj64TJmG1Kk/0bHydLACFIs2cl40eR/SoBPDb6U1S4QoeVqGGTigWERQsu521ZUonOmwFFPSD2+THEhz9iSj3DKImGk5bX0FRfEc+qJ5KWV5dU5A1ey84WU4uFF9Z+4/OU6denrfc8PDpQhyxMLHu9I4D3/YMJSd084RzxOTwMR6kF6Xadkh4yxmynOOUvksscK6NZM5QS5W2QuuL+A8Nz1tJGloRtbJyTKJ9J4LJAoAjRT8zBBHlBaxd/J37XM+zHW6mjDEADocRnfgmbsMujfmCPAEkUQM6jnx3mtIyiYm5i7dGJqUtNhMHhfRHbHmKeZe+kiWcvMi+96XC8W0UN1xUnRnFkLPd4g/cmv9yCtZlmrOncfXMH++vTzwUoYXRTyCmaut5/ScsMwCIj94k+JcdVZxT1rhqIs1Yj/flgbrHlQwZ30l4tTFhMiZmk9N60BscnGWSo8AVsfwSKzXLMHNQcuJUZerouHnwDimFNg6AeEwWVavKuw4yVwIeBRgfM8uyToWw4Ujrp4ilqbAxmAL2SHGnRIZllrTL9HJ1rnhqSooM90ViFUbW4vteKEkH6aHmJhgSmyjOwVNnx/FNKarhBiSEDayLmHzQmKWMs/FxkJU7kmDkfRSIyAnVZHk6/w1KnqNMGc64uMdj+UyDFov3M4sTienu3PMr0bJRGTo8odwHgaZEJhSLLINMrPGwjT2DqMRqoZg1rXLTrISl0Ny3KiybHwQIrT1J4wAHn2oyPT1AjBTYhceCm6FZyCc5GZTUjPizi5tqj3JA8hoQowq0TYGTMwDGGvi1pgxyWmwnkBMb68QFpSeSwdzqxT0YxIB4cjgQpPFoxjdWQ6DSurUK5CnkURNDZEnNgKeYJG79WNSexHEXhNHOlbC5gITGVkygMnKVaanGpMUZFj8BccRf6I6c7xzr7De+OPCKjMjwOtPQLiSkp7nzSQrQ7fmA+sqXfSs1l2gOaUI5n03zsidXtxvaY0bjwR7dWG9CiQTz04uIr4sac3Xy1oBYV7gzxdymcrOQrGKTBoOmIlyUjiC1MXoNCHuKHryURbnJla++Gn9YT3FjA247r5
|
|||
|
|
"text/plain": [
|
|||
|
|
"<PIL.Image.Image image mode=RGB size=512x288>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 43,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"import tensorflow_hub as hub\n",
|
|||
|
|
"hub_model = hub.load('https://tfhub.dev/google/magenta/arbitrary-image-stylization-v1-256/2')\n",
|
|||
|
|
"stylized_image = hub_model(tf.constant(content_image), tf.constant(style_image))[0]\n",
|
|||
|
|
"tensor_to_image(stylized_image)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "GEwZ7FlwrjoZ"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Define content and style representations\n",
|
|||
|
|
"\n",
|
|||
|
|
"Use the intermediate layers of the model to get the *content* and *style* representations of the image. Starting from the network's input layer, the first few layer activations represent low-level features like edges and textures. As you step through the network, the final few layers represent higher-level features—object parts like *wheels* or *eyes*. In this case, you are using the VGG19 network architecture, a pretrained image classification network. These intermediate layers are necessary to define the representation of content and style from the images. For an input image, try to match the corresponding style and content target representations at these intermediate layers.\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "LP_7zrziuiJk"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Load a [VGG19](https://keras.io/api/applications/vgg/#vgg19-function) and test run it on our image to ensure it's used correctly:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 10,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:34:55.337831Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:34:55.337587Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:02.690946Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:02.690299Z"
|
|||
|
|
},
|
|||
|
|
"id": "fMbzrr7BCTq0"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stderr",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"2024-02-23 18:42:24.673107: I external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:454] Loaded cuDNN version 8902\n",
|
|||
|
|
"2024-02-23 18:42:24.773440: I external/local_tsl/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory\n",
|
|||
|
|
"2024-02-23 18:42:25.246989: I external/local_tsl/tsl/platform/default/subprocess.cc:304] Start cannot spawn child process: No such file or directory\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"text/plain": [
|
|||
|
|
"TensorShape([1, 1000])"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 10,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"x = tf.keras.applications.vgg19.preprocess_input(content_image*255)\n",
|
|||
|
|
"x = tf.image.resize(x, (224, 224))\n",
|
|||
|
|
"vgg = tf.keras.applications.VGG19(include_top=True, weights='imagenet')\n",
|
|||
|
|
"prediction_probabilities = vgg(x)\n",
|
|||
|
|
"prediction_probabilities.shape"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 13,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"#vgg.summary()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 12,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"#vgg.get_config()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 11,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:02.694190Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:02.693938Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:02.758402Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:02.757842Z"
|
|||
|
|
},
|
|||
|
|
"id": "1_FyCm0dYnvl"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"text/plain": [
|
|||
|
|
"[('megalith', 0.761035),\n",
|
|||
|
|
" ('king_penguin', 0.20216931),\n",
|
|||
|
|
" ('ram', 0.014038121),\n",
|
|||
|
|
" ('hay', 0.009777568),\n",
|
|||
|
|
" ('bighorn', 0.0025387204)]"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 11,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"predicted_top_5 = tf.keras.applications.vgg19.decode_predictions(prediction_probabilities.numpy())[0]\n",
|
|||
|
|
"[(class_name, prob) for (number, class_name, prob) in predicted_top_5]"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "ljpoYk-0f6HS"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Now load a `VGG19` without the classification head, and list the layer names"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 14,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:02.761485Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:02.761261Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:03.621343Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:03.620659Z"
|
|||
|
|
},
|
|||
|
|
"id": "Yh_AV6220ebD"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"\n",
|
|||
|
|
"input_2\n",
|
|||
|
|
"block1_conv1\n",
|
|||
|
|
"block1_conv2\n",
|
|||
|
|
"block1_pool\n",
|
|||
|
|
"block2_conv1\n",
|
|||
|
|
"block2_conv2\n",
|
|||
|
|
"block2_pool\n",
|
|||
|
|
"block3_conv1\n",
|
|||
|
|
"block3_conv2\n",
|
|||
|
|
"block3_conv3\n",
|
|||
|
|
"block3_conv4\n",
|
|||
|
|
"block3_pool\n",
|
|||
|
|
"block4_conv1\n",
|
|||
|
|
"block4_conv2\n",
|
|||
|
|
"block4_conv3\n",
|
|||
|
|
"block4_conv4\n",
|
|||
|
|
"block4_pool\n",
|
|||
|
|
"block5_conv1\n",
|
|||
|
|
"block5_conv2\n",
|
|||
|
|
"block5_conv3\n",
|
|||
|
|
"block5_conv4\n",
|
|||
|
|
"block5_pool\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"vgg = tf.keras.applications.VGG19(include_top=False, weights='imagenet')\n",
|
|||
|
|
"\n",
|
|||
|
|
"print()\n",
|
|||
|
|
"for layer in vgg.layers:\n",
|
|||
|
|
" print(layer.name)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "Wt-tASys0eJv"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Choose intermediate layers from the network to represent the style and content of the image:\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 15,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:03.624855Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:03.624612Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:03.628342Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:03.627754Z"
|
|||
|
|
},
|
|||
|
|
"id": "ArfX_6iA0WAX"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"content_layers = ['block5_conv2'] \n",
|
|||
|
|
"\n",
|
|||
|
|
"style_layers = ['block1_conv1',\n",
|
|||
|
|
" 'block2_conv1',\n",
|
|||
|
|
" 'block3_conv1', \n",
|
|||
|
|
" 'block4_conv1', \n",
|
|||
|
|
" 'block5_conv1']\n",
|
|||
|
|
"\n",
|
|||
|
|
"num_content_layers = len(content_layers)\n",
|
|||
|
|
"num_style_layers = len(style_layers)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "2o4nSwuN0U3X"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"#### Intermediate layers for style and content\n",
|
|||
|
|
"\n",
|
|||
|
|
"So why do these intermediate outputs within our pretrained image classification network allow us to define style and content representations?\n",
|
|||
|
|
"\n",
|
|||
|
|
"At a high level, in order for a network to perform image classification (which this network has been trained to do), it must understand the image. This requires taking the raw image as input pixels and building an internal representation that converts the raw image pixels into a complex understanding of the features present within the image.\n",
|
|||
|
|
"\n",
|
|||
|
|
"This is also a reason why convolutional neural networks are able to generalize well: they’re able to capture the invariances and defining features within classes (e.g. cats vs. dogs) that are agnostic to background noise and other nuisances. Thus, somewhere between where the raw image is fed into the model and the output classification label, the model serves as a complex feature extractor. By accessing intermediate layers of the model, you're able to describe the content and style of input images."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "Jt3i3RRrJiOX"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Build the model \n",
|
|||
|
|
"\n",
|
|||
|
|
"The networks in `tf.keras.applications` are designed so you can easily extract the intermediate layer values using the Keras functional API.\n",
|
|||
|
|
"\n",
|
|||
|
|
"To define a model using the functional API, specify the inputs and outputs:\n",
|
|||
|
|
"\n",
|
|||
|
|
"`model = Model(inputs, outputs)`\n",
|
|||
|
|
"\n",
|
|||
|
|
"This following function builds a VGG19 model that returns a list of intermediate layer outputs:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 16,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:03.631746Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:03.631221Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:03.635499Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:03.634948Z"
|
|||
|
|
},
|
|||
|
|
"id": "nfec6MuMAbPx"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def vgg_layers(layer_names):\n",
|
|||
|
|
" \"\"\" Creates a VGG model that returns a list of intermediate output values.\"\"\"\n",
|
|||
|
|
" # Load our model. Load pretrained VGG, trained on ImageNet data\n",
|
|||
|
|
" vgg = tf.keras.applications.VGG19(include_top=False, weights='imagenet')\n",
|
|||
|
|
" vgg.trainable = False\n",
|
|||
|
|
" \n",
|
|||
|
|
" outputs = [vgg.get_layer(name).output for name in layer_names]\n",
|
|||
|
|
"\n",
|
|||
|
|
" model = tf.keras.Model([vgg.input], outputs)\n",
|
|||
|
|
" return model"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "jbaIvZf5wWn_"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"And to create the model:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 17,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:03.638264Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:03.638051Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:05.887203Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:05.886443Z"
|
|||
|
|
},
|
|||
|
|
"id": "LkyvPpBHSfVi"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"block1_conv1\n",
|
|||
|
|
" shape: (1, 403, 512, 64)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 743.2428\n",
|
|||
|
|
" mean: 32.208652\n",
|
|||
|
|
"\n",
|
|||
|
|
"block2_conv1\n",
|
|||
|
|
" shape: (1, 201, 256, 128)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 2573.0193\n",
|
|||
|
|
" mean: 134.44252\n",
|
|||
|
|
"\n",
|
|||
|
|
"block3_conv1\n",
|
|||
|
|
" shape: (1, 100, 128, 256)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 8561.582\n",
|
|||
|
|
" mean: 103.2595\n",
|
|||
|
|
"\n",
|
|||
|
|
"block4_conv1\n",
|
|||
|
|
" shape: (1, 50, 64, 512)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 11601.054\n",
|
|||
|
|
" mean: 451.57892\n",
|
|||
|
|
"\n",
|
|||
|
|
"block5_conv1\n",
|
|||
|
|
" shape: (1, 25, 32, 512)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 1913.4247\n",
|
|||
|
|
" mean: 32.27252\n",
|
|||
|
|
"\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"style_extractor = vgg_layers(style_layers)\n",
|
|||
|
|
"style_outputs = style_extractor(style_image*255)\n",
|
|||
|
|
"\n",
|
|||
|
|
"#Look at the statistics of each layer's output\n",
|
|||
|
|
"for name, output in zip(style_layers, style_outputs):\n",
|
|||
|
|
" print(name)\n",
|
|||
|
|
" print(\" shape: \", output.numpy().shape)\n",
|
|||
|
|
" print(\" min: \", output.numpy().min())\n",
|
|||
|
|
" print(\" max: \", output.numpy().max())\n",
|
|||
|
|
" print(\" mean: \", output.numpy().mean())\n",
|
|||
|
|
" print()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "lGUfttK9F8d5"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Calculate style\n",
|
|||
|
|
"\n",
|
|||
|
|
"The content of an image is represented by the values of the intermediate feature maps.\n",
|
|||
|
|
"\n",
|
|||
|
|
"It turns out, the style of an image can be described by the means and correlations across the different feature maps. Calculate a Gram matrix that includes this information by taking the outer product of the feature vector with itself at each location, and averaging that outer product over all locations. This Gram matrix can be calculated for a particular layer as:\n",
|
|||
|
|
"\n",
|
|||
|
|
"$$G^l_{cd} = \\frac{\\sum_{ij} F^l_{ijc}(x)F^l_{ijd}(x)}{IJ}$$\n",
|
|||
|
|
"\n",
|
|||
|
|
"This can be implemented concisely using the `tf.linalg.einsum` function:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 18,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:05.890933Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:05.890331Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:05.895190Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:05.894366Z"
|
|||
|
|
},
|
|||
|
|
"id": "HAy1iGPdoEpZ"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def gram_matrix(input_tensor):\n",
|
|||
|
|
" result = tf.linalg.einsum('bijc,bijd->bcd', input_tensor, input_tensor)\n",
|
|||
|
|
" input_shape = tf.shape(input_tensor)\n",
|
|||
|
|
" num_locations = tf.cast(input_shape[1]*input_shape[2], tf.float32)\n",
|
|||
|
|
" return result/(num_locations)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "pXIUX6czZABh"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Extract style and content\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "1HGHvwlJ1nkn"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Build a model that returns the style and content tensors."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 19,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:05.898313Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:05.898052Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:05.905039Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:05.904179Z"
|
|||
|
|
},
|
|||
|
|
"id": "Sr6QALY-I1ja"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"class StyleContentModel(tf.keras.models.Model):\n",
|
|||
|
|
" def __init__(self, style_layers, content_layers):\n",
|
|||
|
|
" super(StyleContentModel, self).__init__()\n",
|
|||
|
|
" self.vgg = vgg_layers(style_layers + content_layers)\n",
|
|||
|
|
" self.style_layers = style_layers\n",
|
|||
|
|
" self.content_layers = content_layers\n",
|
|||
|
|
" self.num_style_layers = len(style_layers)\n",
|
|||
|
|
" self.vgg.trainable = False\n",
|
|||
|
|
"\n",
|
|||
|
|
" def call(self, inputs):\n",
|
|||
|
|
" \"Expects float input in [0,1]\"\n",
|
|||
|
|
" inputs = inputs*255.0\n",
|
|||
|
|
" preprocessed_input = tf.keras.applications.vgg19.preprocess_input(inputs)\n",
|
|||
|
|
" outputs = self.vgg(preprocessed_input)\n",
|
|||
|
|
" style_outputs, content_outputs = (outputs[:self.num_style_layers],\n",
|
|||
|
|
" outputs[self.num_style_layers:])\n",
|
|||
|
|
"\n",
|
|||
|
|
" style_outputs = [gram_matrix(style_output)\n",
|
|||
|
|
" for style_output in style_outputs]\n",
|
|||
|
|
"\n",
|
|||
|
|
" content_dict = {content_name: value\n",
|
|||
|
|
" for content_name, value\n",
|
|||
|
|
" in zip(self.content_layers, content_outputs)}\n",
|
|||
|
|
"\n",
|
|||
|
|
" style_dict = {style_name: value\n",
|
|||
|
|
" for style_name, value\n",
|
|||
|
|
" in zip(self.style_layers, style_outputs)}\n",
|
|||
|
|
"\n",
|
|||
|
|
" return {'content': content_dict, 'style': style_dict}"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "Xuj1o33t1edl"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"When called on an image, this model returns the gram matrix (style) of the `style_layers` and content of the `content_layers`:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 20,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:05.908129Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:05.907892Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.339329Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.338644Z"
|
|||
|
|
},
|
|||
|
|
"id": "rkjO-DoNDU0A"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Styles:\n",
|
|||
|
|
" block1_conv1\n",
|
|||
|
|
" shape: (1, 64, 64)\n",
|
|||
|
|
" min: 0.022261642\n",
|
|||
|
|
" max: 44157.473\n",
|
|||
|
|
" mean: 355.99878\n",
|
|||
|
|
"\n",
|
|||
|
|
" block2_conv1\n",
|
|||
|
|
" shape: (1, 128, 128)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 74937.36\n",
|
|||
|
|
" mean: 11677.908\n",
|
|||
|
|
"\n",
|
|||
|
|
" block3_conv1\n",
|
|||
|
|
" shape: (1, 256, 256)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 379654.12\n",
|
|||
|
|
" mean: 10265.242\n",
|
|||
|
|
"\n",
|
|||
|
|
" block4_conv1\n",
|
|||
|
|
" shape: (1, 512, 512)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 3333814.5\n",
|
|||
|
|
" mean: 183739.55\n",
|
|||
|
|
"\n",
|
|||
|
|
" block5_conv1\n",
|
|||
|
|
" shape: (1, 512, 512)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 238887.22\n",
|
|||
|
|
" mean: 2304.3767\n",
|
|||
|
|
"\n",
|
|||
|
|
"Contents:\n",
|
|||
|
|
" block5_conv2\n",
|
|||
|
|
" shape: (1, 18, 32, 512)\n",
|
|||
|
|
" min: 0.0\n",
|
|||
|
|
" max: 1443.7827\n",
|
|||
|
|
" mean: 16.3132\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"extractor = StyleContentModel(style_layers, content_layers)\n",
|
|||
|
|
"\n",
|
|||
|
|
"results = extractor(tf.constant(content_image))\n",
|
|||
|
|
"\n",
|
|||
|
|
"print('Styles:')\n",
|
|||
|
|
"for name, output in sorted(results['style'].items()):\n",
|
|||
|
|
" print(\" \", name)\n",
|
|||
|
|
" print(\" shape: \", output.numpy().shape)\n",
|
|||
|
|
" print(\" min: \", output.numpy().min())\n",
|
|||
|
|
" print(\" max: \", output.numpy().max())\n",
|
|||
|
|
" print(\" mean: \", output.numpy().mean())\n",
|
|||
|
|
" print()\n",
|
|||
|
|
"\n",
|
|||
|
|
"print(\"Contents:\")\n",
|
|||
|
|
"for name, output in sorted(results['content'].items()):\n",
|
|||
|
|
" print(\" \", name)\n",
|
|||
|
|
" print(\" shape: \", output.numpy().shape)\n",
|
|||
|
|
" print(\" min: \", output.numpy().min())\n",
|
|||
|
|
" print(\" max: \", output.numpy().max())\n",
|
|||
|
|
" print(\" mean: \", output.numpy().mean())\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "y9r8Lyjb_m0u"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Run gradient descent\n",
|
|||
|
|
"\n",
|
|||
|
|
"With this style and content extractor, you can now implement the style transfer algorithm. Do this by calculating the mean square error for your image's output relative to each target, then take the weighted sum of these losses.\n",
|
|||
|
|
"\n",
|
|||
|
|
"Set your style and content target values:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 21,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.342604Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.342348Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.415356Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.414749Z"
|
|||
|
|
},
|
|||
|
|
"id": "PgkNOnGUFcKa"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"style_targets = extractor(style_image)['style']\n",
|
|||
|
|
"content_targets = extractor(content_image)['content']"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "CNPrpl-e_w9A"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Define a `tf.Variable` to contain the image to optimize. To make this quick, initialize it with the content image (the `tf.Variable` must be the same shape as the content image):"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 22,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.418956Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.418727Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.423759Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.423087Z"
|
|||
|
|
},
|
|||
|
|
"id": "J0vKxF8ZO6G8"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"image = tf.Variable(content_image)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "M6L8ojmn_6rH"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Since this is a float image, define a function to keep the pixel values between 0 and 1:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 23,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.426850Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.426609Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.429855Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.429296Z"
|
|||
|
|
},
|
|||
|
|
"id": "kdgpTJwL_vE2"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def clip_0_1(image):\n",
|
|||
|
|
" return tf.clip_by_value(image, clip_value_min=0.0, clip_value_max=1.0)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "MBU5RFpcAo7W"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Create an optimizer. The paper recommends LBFGS, but Adam works okay, too:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 24,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.432914Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.432699Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.439251Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.438680Z"
|
|||
|
|
},
|
|||
|
|
"id": "r4XZjqUk_5Eu"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"opt = tf.keras.optimizers.Adam(learning_rate=0.02, beta_1=0.99, epsilon=1e-1)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "As-evbBiA2qT"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"To optimize this, use a weighted combination of the two losses to get the total loss:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 25,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.442444Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.442228Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.445095Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.444539Z"
|
|||
|
|
},
|
|||
|
|
"id": "Dt4pxarvA4I4"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"style_weight=1e-2\n",
|
|||
|
|
"content_weight=1e4"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 26,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.447881Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.447669Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.451897Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.451340Z"
|
|||
|
|
},
|
|||
|
|
"id": "0ggx2Na8oROH"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def style_content_loss(outputs):\n",
|
|||
|
|
" style_outputs = outputs['style']\n",
|
|||
|
|
" content_outputs = outputs['content']\n",
|
|||
|
|
" style_loss = tf.add_n([tf.reduce_mean((style_outputs[name]-style_targets[name])**2) \n",
|
|||
|
|
" for name in style_outputs.keys()])\n",
|
|||
|
|
" style_loss *= style_weight / num_style_layers\n",
|
|||
|
|
"\n",
|
|||
|
|
" content_loss = tf.add_n([tf.reduce_mean((content_outputs[name]-content_targets[name])**2) \n",
|
|||
|
|
" for name in content_outputs.keys()])\n",
|
|||
|
|
" content_loss *= content_weight / num_content_layers\n",
|
|||
|
|
" loss = style_loss + content_loss\n",
|
|||
|
|
" return loss"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "vbF2WnP9BI5M"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Use `tf.GradientTape` to update the image."
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 27,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.454809Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.454595Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:08.458356Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:08.457811Z"
|
|||
|
|
},
|
|||
|
|
"id": "0t0umkajFIuh"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"@tf.function()\n",
|
|||
|
|
"def train_step(image):\n",
|
|||
|
|
" with tf.GradientTape() as tape:\n",
|
|||
|
|
" outputs = extractor(image)\n",
|
|||
|
|
" loss = style_content_loss(outputs)\n",
|
|||
|
|
"\n",
|
|||
|
|
" grad = tape.gradient(loss, image)\n",
|
|||
|
|
" opt.apply_gradients([(grad, image)])\n",
|
|||
|
|
" image.assign(clip_0_1(image))"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "5FHMJq4UBRIQ"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Now run a few steps to test:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 28,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:08.461299Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:08.461085Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:35:13.516037Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:35:13.515351Z"
|
|||
|
|
},
|
|||
|
|
"id": "Y542mxi-O2a2"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"name": "stderr",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"2024-02-23 18:43:23.674927: I external/local_xla/xla/service/service.cc:168] XLA service 0x7f88f61b4e70 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
|
|||
|
|
"2024-02-23 18:43:23.674967: I external/local_xla/xla/service/service.cc:176] StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
|
|||
|
|
"2024-02-23 18:43:23.678877: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
|
|||
|
|
"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
|
|||
|
|
"I0000 00:00:1708710203.770091 50541 device_compiler.h:186] Compiled cluster using XLA! This line is logged at most once for the lifetime of the process.\n"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/jpeg": "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
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<PIL.Image.Image image mode=RGB size=512x288>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 28,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"train_step(image)\n",
|
|||
|
|
"train_step(image)\n",
|
|||
|
|
"train_step(image)\n",
|
|||
|
|
"tensor_to_image(image)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "mNzE-mTbBVgY"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Since it's working, perform a longer optimization:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 29,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:35:13.525306Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:35:13.525049Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:26.653022Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:26.652370Z"
|
|||
|
|
},
|
|||
|
|
"id": "rQW1tXYoLbUS"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/jpeg": "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
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<PIL.Image.Image image mode=RGB size=512x288>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Train step: 1000\n",
|
|||
|
|
"Total time: 69.4\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"import time\n",
|
|||
|
|
"start = time.time()\n",
|
|||
|
|
"\n",
|
|||
|
|
"epochs = 10\n",
|
|||
|
|
"steps_per_epoch = 100\n",
|
|||
|
|
"\n",
|
|||
|
|
"step = 0\n",
|
|||
|
|
"for n in range(epochs):\n",
|
|||
|
|
" for m in range(steps_per_epoch):\n",
|
|||
|
|
" step += 1\n",
|
|||
|
|
" train_step(image)\n",
|
|||
|
|
" print(\".\", end='', flush=True)\n",
|
|||
|
|
" display.clear_output(wait=True)\n",
|
|||
|
|
" display.display(tensor_to_image(image))\n",
|
|||
|
|
" print(\"Train step: {}\".format(step))\n",
|
|||
|
|
" \n",
|
|||
|
|
"end = time.time()\n",
|
|||
|
|
"print(\"Total time: {:.1f}\".format(end-start))"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "GWVB3anJMY2v"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Total variation loss\n",
|
|||
|
|
"\n",
|
|||
|
|
"One downside to this basic implementation is that it produces a lot of high frequency artifacts. Decrease these using an explicit regularization term on the high frequency components of the image. In style transfer, this is often called the *total variation loss*:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 30,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:26.664062Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:26.663812Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:26.667988Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:26.667377Z"
|
|||
|
|
},
|
|||
|
|
"id": "7szUUybCQMB3"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def high_pass_x_y(image):\n",
|
|||
|
|
" x_var = image[:, :, 1:, :] - image[:, :, :-1, :]\n",
|
|||
|
|
" y_var = image[:, 1:, :, :] - image[:, :-1, :, :]\n",
|
|||
|
|
"\n",
|
|||
|
|
" return x_var, y_var"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 31,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:26.670924Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:26.670704Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:27.714920Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:27.714172Z"
|
|||
|
|
},
|
|||
|
|
"id": "Atc2oL29PXu_"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<Figure size 1400x1000 with 4 Axes>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"x_deltas, y_deltas = high_pass_x_y(content_image)\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.figure(figsize=(14, 10))\n",
|
|||
|
|
"plt.subplot(2, 2, 1)\n",
|
|||
|
|
"imshow(clip_0_1(2*y_deltas+0.5), \"Horizontal Deltas: Original\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.subplot(2, 2, 2)\n",
|
|||
|
|
"imshow(clip_0_1(2*x_deltas+0.5), \"Vertical Deltas: Original\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"x_deltas, y_deltas = high_pass_x_y(image)\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.subplot(2, 2, 3)\n",
|
|||
|
|
"imshow(clip_0_1(2*y_deltas+0.5), \"Horizontal Deltas: Styled\")\n",
|
|||
|
|
"\n",
|
|||
|
|
"plt.subplot(2, 2, 4)\n",
|
|||
|
|
"imshow(clip_0_1(2*x_deltas+0.5), \"Vertical Deltas: Styled\")"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "lqHElVgBkgkz"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"This shows how the high frequency components have increased.\n",
|
|||
|
|
"\n",
|
|||
|
|
"Also, this high frequency component is basically an edge-detector. You can get similar output from the Sobel edge detector, for example:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 32,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:27.743198Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:27.742554Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.398290Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.397633Z"
|
|||
|
|
},
|
|||
|
|
"id": "HyvqCiywiUfL"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<Figure size 1400x1000 with 2 Axes>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"plt.figure(figsize=(14, 10))\n",
|
|||
|
|
"\n",
|
|||
|
|
"sobel = tf.image.sobel_edges(content_image)\n",
|
|||
|
|
"plt.subplot(1, 2, 1)\n",
|
|||
|
|
"imshow(clip_0_1(sobel[..., 0]/4+0.5), \"Horizontal Sobel-edges\")\n",
|
|||
|
|
"plt.subplot(1, 2, 2)\n",
|
|||
|
|
"imshow(clip_0_1(sobel[..., 1]/4+0.5), \"Vertical Sobel-edges\")"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "vv5bKlSDnPP7"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"The regularization loss associated with this is the sum of the squares of the values:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 33,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.410709Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.410447Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.414372Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.413690Z"
|
|||
|
|
},
|
|||
|
|
"id": "mP-92lXMIYPn"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"def total_variation_loss(image):\n",
|
|||
|
|
" x_deltas, y_deltas = high_pass_x_y(image)\n",
|
|||
|
|
" return tf.reduce_sum(tf.abs(x_deltas)) + tf.reduce_sum(tf.abs(y_deltas))"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 34,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.417541Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.417005Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.426670Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.426108Z"
|
|||
|
|
},
|
|||
|
|
"id": "s4OYBUX2KQ25"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"text/plain": [
|
|||
|
|
"56826.945"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 34,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"total_variation_loss(image).numpy()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "pu2hJ8zOKMc1"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"That demonstrated what it does. But there's no need to implement it yourself, TensorFlow includes a standard implementation:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 35,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.430127Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.429584Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.436614Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.436039Z"
|
|||
|
|
},
|
|||
|
|
"id": "YQjWW04NKLfJ"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"text/plain": [
|
|||
|
|
"array([56826.945], dtype=float32)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 35,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"tf.image.total_variation(image).numpy()"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "nTessd-DCdcC"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Re-run the optimization\n",
|
|||
|
|
"\n",
|
|||
|
|
"Choose a weight for the `total_variation_loss`:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 36,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.439704Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.439220Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.442092Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.441561Z"
|
|||
|
|
},
|
|||
|
|
"id": "tGeRLD4GoAd4"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"total_variation_weight=30"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "kG1-T4kJsoAv"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Now include it in the `train_step` function:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 37,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.445174Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.444649Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.452936Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.452367Z"
|
|||
|
|
},
|
|||
|
|
"id": "BzmfcyyYUyWq"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"@tf.function()\n",
|
|||
|
|
"def train_step(image):\n",
|
|||
|
|
" with tf.GradientTape() as tape:\n",
|
|||
|
|
" outputs = extractor(image)\n",
|
|||
|
|
" loss = style_content_loss(outputs)\n",
|
|||
|
|
" loss += total_variation_weight*tf.image.total_variation(image)\n",
|
|||
|
|
"\n",
|
|||
|
|
" grad = tape.gradient(loss, image)\n",
|
|||
|
|
" opt.apply_gradients([(grad, image)])\n",
|
|||
|
|
" image.assign(clip_0_1(image))"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "lcLWBQChsutQ"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"Reinitialize the image-variable and the optimizer:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 38,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.455824Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.455522Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:36:28.460788Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:36:28.460227Z"
|
|||
|
|
},
|
|||
|
|
"id": "a-dPRr8BqexB"
|
|||
|
|
},
|
|||
|
|
"outputs": [],
|
|||
|
|
"source": [
|
|||
|
|
"opt = tf.keras.optimizers.Adam(learning_rate=0.02, beta_1=0.99, epsilon=1e-1)\n",
|
|||
|
|
"image = tf.Variable(content_image)"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "BEflRstmtGBu"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"And run the optimization:"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 39,
|
|||
|
|
"metadata": {
|
|||
|
|
"execution": {
|
|||
|
|
"iopub.execute_input": "2023-11-16T03:36:28.463940Z",
|
|||
|
|
"iopub.status.busy": "2023-11-16T03:36:28.463420Z",
|
|||
|
|
"iopub.status.idle": "2023-11-16T03:37:43.245043Z",
|
|||
|
|
"shell.execute_reply": "2023-11-16T03:37:43.244317Z"
|
|||
|
|
},
|
|||
|
|
"id": "q3Cc3bLtoOWy"
|
|||
|
|
},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"image/jpeg": "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
|
|||
|
|
"image/png": "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
|
|||
|
|
"text/plain": [
|
|||
|
|
"<PIL.Image.Image image mode=RGB size=512x288>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "display_data"
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"name": "stdout",
|
|||
|
|
"output_type": "stream",
|
|||
|
|
"text": [
|
|||
|
|
"Train step: 1000\n",
|
|||
|
|
"Total time: 71.3\n"
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"import time\n",
|
|||
|
|
"start = time.time()\n",
|
|||
|
|
"\n",
|
|||
|
|
"epochs = 10\n",
|
|||
|
|
"steps_per_epoch = 100\n",
|
|||
|
|
"\n",
|
|||
|
|
"step = 0\n",
|
|||
|
|
"for n in range(epochs):\n",
|
|||
|
|
" for m in range(steps_per_epoch):\n",
|
|||
|
|
" step += 1\n",
|
|||
|
|
" train_step(image)\n",
|
|||
|
|
" print(\".\", end='', flush=True)\n",
|
|||
|
|
" display.clear_output(wait=True)\n",
|
|||
|
|
" display.display(tensor_to_image(image))\n",
|
|||
|
|
" print(\"Train step: {}\".format(step))\n",
|
|||
|
|
"\n",
|
|||
|
|
"end = time.time()\n",
|
|||
|
|
"print(\"Total time: {:.1f}\".format(end-start))"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "code",
|
|||
|
|
"execution_count": 44,
|
|||
|
|
"metadata": {},
|
|||
|
|
"outputs": [
|
|||
|
|
{
|
|||
|
|
"data": {
|
|||
|
|
"text/plain": [
|
|||
|
|
"<tf.Variable 'Variable:0' shape=(1, 422, 512, 3) dtype=float32, numpy=\n",
|
|||
|
|
"array([[[[5.2192467e-01, 3.5730761e-01, 4.3012524e-01],\n",
|
|||
|
|
" [5.2258748e-01, 3.5042602e-01, 4.2847326e-01],\n",
|
|||
|
|
" [5.2142912e-01, 3.3703288e-01, 4.6795484e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [3.5259786e-01, 2.0069836e-01, 0.0000000e+00],\n",
|
|||
|
|
" [3.7858683e-01, 8.8987187e-02, 0.0000000e+00],\n",
|
|||
|
|
" [3.7882906e-01, 8.8669561e-02, 0.0000000e+00]],\n",
|
|||
|
|
"\n",
|
|||
|
|
" [[5.2074802e-01, 3.5717541e-01, 4.3073651e-01],\n",
|
|||
|
|
" [5.2213031e-01, 3.5292444e-01, 4.2814627e-01],\n",
|
|||
|
|
" [5.1997304e-01, 3.3482769e-01, 4.7205076e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [3.5210976e-01, 2.1672039e-01, 0.0000000e+00],\n",
|
|||
|
|
" [3.9774185e-01, 9.4535254e-02, 0.0000000e+00],\n",
|
|||
|
|
" [4.0475607e-01, 9.9769264e-02, 0.0000000e+00]],\n",
|
|||
|
|
"\n",
|
|||
|
|
" [[5.2096248e-01, 4.2476922e-01, 5.6723040e-01],\n",
|
|||
|
|
" [5.1835060e-01, 3.7873101e-01, 5.5380058e-01],\n",
|
|||
|
|
" [5.1772678e-01, 3.3783364e-01, 5.5342436e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [3.5153681e-01, 2.9214185e-01, 0.0000000e+00],\n",
|
|||
|
|
" [4.0174142e-01, 2.9518455e-01, 0.0000000e+00],\n",
|
|||
|
|
" [4.1466421e-01, 2.9645365e-01, 9.2595583e-06]],\n",
|
|||
|
|
"\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
"\n",
|
|||
|
|
" [[7.7451438e-01, 6.1107898e-01, 4.9147427e-01],\n",
|
|||
|
|
" [7.9484630e-01, 5.3058982e-01, 4.9114281e-01],\n",
|
|||
|
|
" [5.5543298e-01, 5.2111608e-01, 4.6575382e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [4.6427000e-01, 5.5372471e-01, 5.6248754e-01],\n",
|
|||
|
|
" [2.1076988e-01, 3.1807196e-01, 5.5500793e-01],\n",
|
|||
|
|
" [2.0806797e-01, 3.1826279e-01, 5.5626237e-01]],\n",
|
|||
|
|
"\n",
|
|||
|
|
" [[5.5564922e-01, 5.2309209e-01, 4.7513846e-01],\n",
|
|||
|
|
" [5.5392897e-01, 5.2152395e-01, 4.7580311e-01],\n",
|
|||
|
|
" [5.5371583e-01, 5.1829290e-01, 4.6731946e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [4.6132144e-01, 5.4868215e-01, 5.6376553e-01],\n",
|
|||
|
|
" [1.7865986e-01, 2.9668301e-01, 5.6151551e-01],\n",
|
|||
|
|
" [1.6780509e-01, 2.9647833e-01, 5.6041163e-01]],\n",
|
|||
|
|
"\n",
|
|||
|
|
" [[5.5324817e-01, 5.2326709e-01, 4.7086918e-01],\n",
|
|||
|
|
" [5.5364186e-01, 5.2139556e-01, 4.7007531e-01],\n",
|
|||
|
|
" [5.5321431e-01, 5.1366639e-01, 4.7087216e-01],\n",
|
|||
|
|
" ...,\n",
|
|||
|
|
" [4.6121120e-01, 5.4781228e-01, 5.6475556e-01],\n",
|
|||
|
|
" [1.7181176e-01, 2.9641071e-01, 5.6171346e-01],\n",
|
|||
|
|
" [1.7033017e-01, 2.9690620e-01, 5.5929840e-01]]]], dtype=float32)>"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
"execution_count": 44,
|
|||
|
|
"metadata": {},
|
|||
|
|
"output_type": "execute_result"
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"source": [
|
|||
|
|
"image"
|
|||
|
|
]
|
|||
|
|
},
|
|||
|
|
{
|
|||
|
|
"cell_type": "markdown",
|
|||
|
|
"metadata": {
|
|||
|
|
"id": "tNlwRXagxQZk"
|
|||
|
|
},
|
|||
|
|
"source": [
|
|||
|
|
"## Learn more\n",
|
|||
|
|
"\n",
|
|||
|
|
"This tutorial demonstrates the original style-transfer algorithm. For a simple application of style transfer check out this [tutorial](https://www.tensorflow.org/hub/tutorials/tf2_arbitrary_image_stylization) to learn more about how to use the arbitrary image style transfer model from [TensorFlow Hub](https://tfhub.dev)."
|
|||
|
|
]
|
|||
|
|
}
|
|||
|
|
],
|
|||
|
|
"metadata": {
|
|||
|
|
"accelerator": "GPU",
|
|||
|
|
"colab": {
|
|||
|
|
"collapsed_sections": [],
|
|||
|
|
"name": "style_transfer.ipynb",
|
|||
|
|
"toc_visible": true
|
|||
|
|
},
|
|||
|
|
"kernelspec": {
|
|||
|
|
"display_name": "Python 3",
|
|||
|
|
"name": "python3"
|
|||
|
|
},
|
|||
|
|
"language_info": {
|
|||
|
|
"codemirror_mode": {
|
|||
|
|
"name": "ipython",
|
|||
|
|
"version": 3
|
|||
|
|
},
|
|||
|
|
"file_extension": ".py",
|
|||
|
|
"mimetype": "text/x-python",
|
|||
|
|
"name": "python",
|
|||
|
|
"nbconvert_exporter": "python",
|
|||
|
|
"pygments_lexer": "ipython3",
|
|||
|
|
"version": "3.10.12"
|
|||
|
|
}
|
|||
|
|
},
|
|||
|
|
"nbformat": 4,
|
|||
|
|
"nbformat_minor": 0
|
|||
|
|
}
|