216 KiB
216 KiB
Zadanie 8 - Alexnet + Dropout & BatchRegularization
Aleksandra Jonas, Aleksandra Gronowska, Iwona Christop
Przygotowanie danych
from IPython.display import Image, displayimport sys
import subprocess
import pkg_resources
import numpy as np
required = { 'scikit-image'}
installed = {pkg.key for pkg in pkg_resources.working_set}
missing = required - installed
# Alexnet requires images to be of dim = (227, 227, 3)
newSize = (227,227)
if missing:
python = sys.executable
subprocess.check_call([python, '-m', 'pip', 'install', *missing], stdout=subprocess.DEVNULL)
def load_train_data(input_dir):
import numpy as np
import pandas as pd
import os
from skimage.io import imread
import cv2 as cv
from pathlib import Path
import random
from shutil import copyfile, rmtree
import json
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib
image_dir = Path(input_dir)
categories_name = []
for file in os.listdir(image_dir):
d = os.path.join(image_dir, file)
if os.path.isdir(d):
categories_name.append(file)
folders = [directory for directory in image_dir.iterdir() if directory.is_dir()]
train_img = []
categories_count=[]
labels=[]
for i, direc in enumerate(folders):
count = 0
for obj in direc.iterdir():
if os.path.isfile(obj) and os.path.basename(os.path.normpath(obj)) != 'desktop.ini':
labels.append(os.path.basename(os.path.normpath(direc)))
count += 1
img = imread(obj)#zwraca ndarry postaci xSize x ySize x colorDepth
img = img[:, :, :3]
img = cv.resize(img, newSize, interpolation=cv.INTER_AREA)# zwraca ndarray
img = img / 255 #normalizacja
train_img.append(img)
categories_count.append(count)
X={}
X["values"] = np.array(train_img)
X["categories_name"] = categories_name
X["categories_count"] = categories_count
X["labels"]=labels
return X
def load_test_data(input_dir):
import numpy as np
import pandas as pd
import os
from skimage.io import imread
import cv2 as cv
from pathlib import Path
import random
from shutil import copyfile, rmtree
import json
import seaborn as sns
import matplotlib.pyplot as plt
import matplotlib
image_path = Path(input_dir)
labels_path = image_path.parents[0] / 'test_labels.json'
jsonString = labels_path.read_text()
objects = json.loads(jsonString)
categories_name = []
categories_count=[]
count = 0
c = objects[0]['value']
for e in objects:
if e['value'] != c:
categories_count.append(count)
c = e['value']
count = 1
else:
count += 1
if not e['value'] in categories_name:
categories_name.append(e['value'])
categories_count.append(count)
test_img = []
labels=[]
for e in objects:
p = image_path / e['filename']
img = imread(p)#zwraca ndarry postaci xSize x ySize x colorDepth
img = img[:, :, :3]
img = cv.resize(img, newSize, interpolation=cv.INTER_AREA)# zwraca ndarray
img = img / 255#normalizacja
test_img.append(img)
labels.append(e['value'])
X={}
X["values"] = np.array(test_img)
X["categories_name"] = categories_name
X["categories_count"] = categories_count
X["labels"]=labels
return X# Data load
data_train = load_train_data("./train_test_sw/train_sw")
values_train = data_train['values']
labels_train = data_train['labels']
data_test = load_test_data("./train_test_sw/test_sw")
X_test = data_test['values']
y_test = data_test['labels']from sklearn.model_selection import train_test_splitX_train, X_validate, y_train, y_validate = train_test_split(values_train, labels_train, test_size=0.2, random_state=42)from sklearn.preprocessing import LabelEncoderclass_le = LabelEncoder()
y_train_enc = class_le.fit_transform(y_train)
y_validate_enc = class_le.fit_transform(y_validate)
y_test_enc = class_le.fit_transform(y_test)import tensorflow as tftrain_ds = tf.data.Dataset.from_tensor_slices((X_train, y_train_enc))
validation_ds = tf.data.Dataset.from_tensor_slices((X_validate, y_validate_enc))
test_ds = tf.data.Dataset.from_tensor_slices((X_test, y_test_enc))train_ds_size = tf.data.experimental.cardinality(train_ds).numpy()
test_ds_size = tf.data.experimental.cardinality(test_ds).numpy()
validation_ds_size = tf.data.experimental.cardinality(validation_ds).numpy()
print("Training data size:", train_ds_size)
print("Test data size:", test_ds_size)
print("Validation data size:", validation_ds_size)Training data size: 820 Test data size: 259 Validation data size: 206
train_ds = (train_ds
.shuffle(buffer_size=train_ds_size)
.batch(batch_size=32, drop_remainder=True))
test_ds = (test_ds
.shuffle(buffer_size=train_ds_size)
.batch(batch_size=32, drop_remainder=True))
validation_ds = (validation_ds
.shuffle(buffer_size=train_ds_size)
.batch(batch_size=32, drop_remainder=True))from tensorflow import keras
import os
import timeroot_logdir = os.path.join(os.curdir, "logs\\\\fit\\\\")
def get_run_logdir():
run_id = time.strftime("run_%Y_%m_%d-%H_%M_%S")
return os.path.join(root_logdir, run_id)
run_logdir = get_run_logdir()
tensorboard_cb = keras.callbacks.TensorBoard(run_logdir)Dropout
Do warstw spłaszczonych
model_flat_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(10, activation='softmax')
])model_flat_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_flat_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 55, 55, 96) 34944
max_pooling2d (MaxPooling2D (None, 27, 27, 96) 0
)
conv2d_1 (Conv2D) (None, 27, 27, 256) 614656
max_pooling2d_1 (MaxPooling (None, 13, 13, 256) 0
2D)
conv2d_2 (Conv2D) (None, 13, 13, 384) 885120
conv2d_3 (Conv2D) (None, 13, 13, 384) 1327488
conv2d_4 (Conv2D) (None, 13, 13, 256) 884992
max_pooling2d_2 (MaxPooling (None, 6, 6, 256) 0
2D)
flatten (Flatten) (None, 9216) 0
dense (Dense) (None, 4096) 37752832
dropout (Dropout) (None, 4096) 0
dense_1 (Dense) (None, 4096) 16781312
dropout_1 (Dropout) (None, 4096) 0
dense_2 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_flat_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 39s 1s/step - loss: 2.2584 - accuracy: 0.2000 - val_loss: 2.1890 - val_accuracy: 0.3073 Epoch 2/100 25/25 [==============================] - 33s 1s/step - loss: 1.9739 - accuracy: 0.2275 - val_loss: 1.6961 - val_accuracy: 0.1875 Epoch 3/100 25/25 [==============================] - 33s 1s/step - loss: 1.6904 - accuracy: 0.2288 - val_loss: 1.6021 - val_accuracy: 0.2604 Epoch 4/100 25/25 [==============================] - 34s 1s/step - loss: 1.6571 - accuracy: 0.2138 - val_loss: 1.5939 - val_accuracy: 0.3333 Epoch 5/100 25/25 [==============================] - 34s 1s/step - loss: 1.6340 - accuracy: 0.2400 - val_loss: 1.5403 - val_accuracy: 0.3438 Epoch 6/100 25/25 [==============================] - 34s 1s/step - loss: 1.6254 - accuracy: 0.2650 - val_loss: 1.5925 - val_accuracy: 0.2917 Epoch 7/100 25/25 [==============================] - 34s 1s/step - loss: 1.6075 - accuracy: 0.2600 - val_loss: 1.5318 - val_accuracy: 0.3698 Epoch 8/100 25/25 [==============================] - 34s 1s/step - loss: 1.5569 - accuracy: 0.3338 - val_loss: 1.5195 - val_accuracy: 0.4167 Epoch 9/100 25/25 [==============================] - 34s 1s/step - loss: 1.5345 - accuracy: 0.3425 - val_loss: 1.5741 - val_accuracy: 0.2917 Epoch 10/100 25/25 [==============================] - 34s 1s/step - loss: 1.5055 - accuracy: 0.3500 - val_loss: 1.3982 - val_accuracy: 0.4115 Epoch 11/100 25/25 [==============================] - 34s 1s/step - loss: 1.4744 - accuracy: 0.3600 - val_loss: 1.5340 - val_accuracy: 0.3854 Epoch 12/100 25/25 [==============================] - 34s 1s/step - loss: 1.4548 - accuracy: 0.3913 - val_loss: 1.4387 - val_accuracy: 0.4115 Epoch 13/100 25/25 [==============================] - 34s 1s/step - loss: 1.4088 - accuracy: 0.4038 - val_loss: 1.4665 - val_accuracy: 0.4323 Epoch 14/100 25/25 [==============================] - 34s 1s/step - loss: 1.3404 - accuracy: 0.4437 - val_loss: 1.3196 - val_accuracy: 0.5052 Epoch 15/100 25/25 [==============================] - 37s 1s/step - loss: 1.3122 - accuracy: 0.4512 - val_loss: 1.2624 - val_accuracy: 0.5052 Epoch 16/100 25/25 [==============================] - 37s 1s/step - loss: 1.2144 - accuracy: 0.4925 - val_loss: 1.1976 - val_accuracy: 0.5521 Epoch 17/100 25/25 [==============================] - 36s 1s/step - loss: 1.1543 - accuracy: 0.5000 - val_loss: 1.1166 - val_accuracy: 0.5104 Epoch 18/100 25/25 [==============================] - 34s 1s/step - loss: 1.1334 - accuracy: 0.5100 - val_loss: 1.3203 - val_accuracy: 0.4635 Epoch 19/100 25/25 [==============================] - 34s 1s/step - loss: 1.1212 - accuracy: 0.5288 - val_loss: 1.1281 - val_accuracy: 0.5208 Epoch 20/100 25/25 [==============================] - 34s 1s/step - loss: 1.0779 - accuracy: 0.5250 - val_loss: 1.1841 - val_accuracy: 0.5365 Epoch 21/100 25/25 [==============================] - 35s 1s/step - loss: 1.0472 - accuracy: 0.5300 - val_loss: 1.0747 - val_accuracy: 0.5677 Epoch 22/100 25/25 [==============================] - 32s 1s/step - loss: 1.0487 - accuracy: 0.5337 - val_loss: 1.1981 - val_accuracy: 0.5312 Epoch 23/100 25/25 [==============================] - 35s 1s/step - loss: 1.0377 - accuracy: 0.5288 - val_loss: 1.1231 - val_accuracy: 0.5312 Epoch 24/100 25/25 [==============================] - 34s 1s/step - loss: 1.0117 - accuracy: 0.5600 - val_loss: 1.1580 - val_accuracy: 0.5208 Epoch 25/100 25/25 [==============================] - 33s 1s/step - loss: 0.9578 - accuracy: 0.5813 - val_loss: 1.1586 - val_accuracy: 0.5052 Epoch 26/100 25/25 [==============================] - 34s 1s/step - loss: 0.9469 - accuracy: 0.5913 - val_loss: 1.0026 - val_accuracy: 0.5625 Epoch 27/100 25/25 [==============================] - 33s 1s/step - loss: 0.9810 - accuracy: 0.5600 - val_loss: 0.9896 - val_accuracy: 0.5781 Epoch 28/100 25/25 [==============================] - 34s 1s/step - loss: 0.9239 - accuracy: 0.5825 - val_loss: 1.3049 - val_accuracy: 0.5312 Epoch 29/100 25/25 [==============================] - 36s 1s/step - loss: 0.9495 - accuracy: 0.5625 - val_loss: 0.9832 - val_accuracy: 0.5990 Epoch 30/100 25/25 [==============================] - 37s 1s/step - loss: 0.9322 - accuracy: 0.5725 - val_loss: 1.0204 - val_accuracy: 0.5677 Epoch 31/100 25/25 [==============================] - 36s 1s/step - loss: 0.9098 - accuracy: 0.5925 - val_loss: 1.0027 - val_accuracy: 0.5208 Epoch 32/100 25/25 [==============================] - 36s 1s/step - loss: 0.9288 - accuracy: 0.6050 - val_loss: 1.0639 - val_accuracy: 0.5625 Epoch 33/100 25/25 [==============================] - 37s 1s/step - loss: 0.8865 - accuracy: 0.6000 - val_loss: 0.9128 - val_accuracy: 0.6250 Epoch 34/100 25/25 [==============================] - 37s 1s/step - loss: 0.8653 - accuracy: 0.6137 - val_loss: 0.9095 - val_accuracy: 0.6146 Epoch 35/100 25/25 [==============================] - 39s 2s/step - loss: 0.8462 - accuracy: 0.6250 - val_loss: 1.0470 - val_accuracy: 0.5625 Epoch 36/100 25/25 [==============================] - 38s 2s/step - loss: 0.8485 - accuracy: 0.6012 - val_loss: 1.0561 - val_accuracy: 0.5312 Epoch 37/100 25/25 [==============================] - 34s 1s/step - loss: 0.8349 - accuracy: 0.6363 - val_loss: 0.9282 - val_accuracy: 0.6198 Epoch 38/100 25/25 [==============================] - 34s 1s/step - loss: 0.8450 - accuracy: 0.6137 - val_loss: 0.9833 - val_accuracy: 0.5885 Epoch 39/100 25/25 [==============================] - 34s 1s/step - loss: 0.8209 - accuracy: 0.6288 - val_loss: 1.0049 - val_accuracy: 0.5729 Epoch 40/100 25/25 [==============================] - 34s 1s/step - loss: 0.8213 - accuracy: 0.6237 - val_loss: 0.8432 - val_accuracy: 0.6406 Epoch 41/100 25/25 [==============================] - 36s 1s/step - loss: 0.7931 - accuracy: 0.6413 - val_loss: 1.0859 - val_accuracy: 0.5469 Epoch 42/100 25/25 [==============================] - 34s 1s/step - loss: 0.7652 - accuracy: 0.6612 - val_loss: 0.9302 - val_accuracy: 0.6094 Epoch 43/100 25/25 [==============================] - 34s 1s/step - loss: 0.7603 - accuracy: 0.6562 - val_loss: 0.8186 - val_accuracy: 0.6458 Epoch 44/100 25/25 [==============================] - 34s 1s/step - loss: 0.7221 - accuracy: 0.7125 - val_loss: 0.9047 - val_accuracy: 0.6458 Epoch 45/100 25/25 [==============================] - 33s 1s/step - loss: 0.7501 - accuracy: 0.6675 - val_loss: 0.8114 - val_accuracy: 0.6562 Epoch 46/100 25/25 [==============================] - 34s 1s/step - loss: 0.7169 - accuracy: 0.6925 - val_loss: 0.6990 - val_accuracy: 0.7031 Epoch 47/100 25/25 [==============================] - 35s 1s/step - loss: 0.6610 - accuracy: 0.7212 - val_loss: 0.7242 - val_accuracy: 0.6927 Epoch 48/100 25/25 [==============================] - 34s 1s/step - loss: 0.6846 - accuracy: 0.7075 - val_loss: 0.8100 - val_accuracy: 0.6510 Epoch 49/100 25/25 [==============================] - 35s 1s/step - loss: 0.6308 - accuracy: 0.7300 - val_loss: 0.7077 - val_accuracy: 0.6771 Epoch 50/100 25/25 [==============================] - 36s 1s/step - loss: 0.5996 - accuracy: 0.7425 - val_loss: 0.6889 - val_accuracy: 0.7031 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 0.6341 - accuracy: 0.7575 - val_loss: 0.6953 - val_accuracy: 0.7344 Epoch 52/100 25/25 [==============================] - 36s 1s/step - loss: 0.5816 - accuracy: 0.7525 - val_loss: 0.6998 - val_accuracy: 0.7031 Epoch 53/100 25/25 [==============================] - 36s 1s/step - loss: 0.5872 - accuracy: 0.7650 - val_loss: 0.7242 - val_accuracy: 0.7188 Epoch 54/100 25/25 [==============================] - 36s 1s/step - loss: 0.5597 - accuracy: 0.7812 - val_loss: 0.8982 - val_accuracy: 0.6719 Epoch 55/100 25/25 [==============================] - 38s 2s/step - loss: 0.5644 - accuracy: 0.7825 - val_loss: 0.6864 - val_accuracy: 0.7396 Epoch 56/100 25/25 [==============================] - 36s 1s/step - loss: 0.5228 - accuracy: 0.7763 - val_loss: 0.6602 - val_accuracy: 0.7031 Epoch 57/100 25/25 [==============================] - 35s 1s/step - loss: 0.4987 - accuracy: 0.7887 - val_loss: 0.6512 - val_accuracy: 0.7812 Epoch 58/100 25/25 [==============================] - 35s 1s/step - loss: 0.5320 - accuracy: 0.7862 - val_loss: 0.6612 - val_accuracy: 0.7604 Epoch 59/100 25/25 [==============================] - 33s 1s/step - loss: 0.5273 - accuracy: 0.7775 - val_loss: 0.6148 - val_accuracy: 0.7708 Epoch 60/100 25/25 [==============================] - 34s 1s/step - loss: 0.5277 - accuracy: 0.7875 - val_loss: 0.6853 - val_accuracy: 0.7448 Epoch 61/100 25/25 [==============================] - 34s 1s/step - loss: 0.4958 - accuracy: 0.7975 - val_loss: 0.7478 - val_accuracy: 0.7188 Epoch 62/100 25/25 [==============================] - 36s 1s/step - loss: 0.4867 - accuracy: 0.7900 - val_loss: 0.5890 - val_accuracy: 0.7188 Epoch 63/100 25/25 [==============================] - 34s 1s/step - loss: 0.4677 - accuracy: 0.8050 - val_loss: 0.7454 - val_accuracy: 0.6927 Epoch 64/100 25/25 [==============================] - 33s 1s/step - loss: 0.4793 - accuracy: 0.8225 - val_loss: 0.6097 - val_accuracy: 0.7656 Epoch 65/100 25/25 [==============================] - 34s 1s/step - loss: 0.4755 - accuracy: 0.8188 - val_loss: 0.6033 - val_accuracy: 0.7604 Epoch 66/100 25/25 [==============================] - 33s 1s/step - loss: 0.4239 - accuracy: 0.8263 - val_loss: 0.6168 - val_accuracy: 0.7865 Epoch 67/100 25/25 [==============================] - 33s 1s/step - loss: 0.4531 - accuracy: 0.8200 - val_loss: 0.5624 - val_accuracy: 0.7812 Epoch 68/100 25/25 [==============================] - 33s 1s/step - loss: 0.4397 - accuracy: 0.8388 - val_loss: 0.5696 - val_accuracy: 0.8229 Epoch 69/100 25/25 [==============================] - 33s 1s/step - loss: 0.4543 - accuracy: 0.8087 - val_loss: 0.6307 - val_accuracy: 0.7865 Epoch 70/100 25/25 [==============================] - 37s 1s/step - loss: 0.3984 - accuracy: 0.8438 - val_loss: 0.6143 - val_accuracy: 0.8125 Epoch 71/100 25/25 [==============================] - 36s 1s/step - loss: 0.3903 - accuracy: 0.8525 - val_loss: 0.6269 - val_accuracy: 0.7656 Epoch 72/100 25/25 [==============================] - 34s 1s/step - loss: 0.3671 - accuracy: 0.8525 - val_loss: 0.5292 - val_accuracy: 0.7917 Epoch 73/100 25/25 [==============================] - 33s 1s/step - loss: 0.4349 - accuracy: 0.8175 - val_loss: 0.6612 - val_accuracy: 0.7552 Epoch 74/100 25/25 [==============================] - 33s 1s/step - loss: 0.3605 - accuracy: 0.8712 - val_loss: 0.8456 - val_accuracy: 0.6615 Epoch 75/100 25/25 [==============================] - 34s 1s/step - loss: 0.4310 - accuracy: 0.8300 - val_loss: 0.6901 - val_accuracy: 0.7344 Epoch 76/100 25/25 [==============================] - 35s 1s/step - loss: 0.4062 - accuracy: 0.8263 - val_loss: 0.5469 - val_accuracy: 0.8021 Epoch 77/100 25/25 [==============================] - 34s 1s/step - loss: 0.3645 - accuracy: 0.8612 - val_loss: 0.5918 - val_accuracy: 0.7656 Epoch 78/100 25/25 [==============================] - 33s 1s/step - loss: 0.3444 - accuracy: 0.8600 - val_loss: 0.5622 - val_accuracy: 0.7969 Epoch 79/100 25/25 [==============================] - 34s 1s/step - loss: 0.3611 - accuracy: 0.8562 - val_loss: 0.6412 - val_accuracy: 0.7760 Epoch 80/100 25/25 [==============================] - 33s 1s/step - loss: 0.3853 - accuracy: 0.8475 - val_loss: 0.6283 - val_accuracy: 0.7969 Epoch 81/100 25/25 [==============================] - 35s 1s/step - loss: 0.3076 - accuracy: 0.8763 - val_loss: 0.6677 - val_accuracy: 0.7500 Epoch 82/100 25/25 [==============================] - 37s 1s/step - loss: 0.3204 - accuracy: 0.8813 - val_loss: 0.6045 - val_accuracy: 0.7865 Epoch 83/100 25/25 [==============================] - 34s 1s/step - loss: 0.3183 - accuracy: 0.8763 - val_loss: 0.6581 - val_accuracy: 0.8281 Epoch 84/100 25/25 [==============================] - 34s 1s/step - loss: 0.2913 - accuracy: 0.8863 - val_loss: 0.6067 - val_accuracy: 0.8073 Epoch 85/100 25/25 [==============================] - 34s 1s/step - loss: 0.3194 - accuracy: 0.8662 - val_loss: 0.6241 - val_accuracy: 0.8281 Epoch 86/100 25/25 [==============================] - 33s 1s/step - loss: 0.4045 - accuracy: 0.8487 - val_loss: 0.6254 - val_accuracy: 0.8073 Epoch 87/100 25/25 [==============================] - 35s 1s/step - loss: 0.2853 - accuracy: 0.8863 - val_loss: 0.6204 - val_accuracy: 0.7917 Epoch 88/100 25/25 [==============================] - 36s 1s/step - loss: 0.2844 - accuracy: 0.8737 - val_loss: 0.5187 - val_accuracy: 0.8177 Epoch 89/100 25/25 [==============================] - 34s 1s/step - loss: 0.2622 - accuracy: 0.8913 - val_loss: 0.7985 - val_accuracy: 0.7656 Epoch 90/100 25/25 [==============================] - 33s 1s/step - loss: 0.3699 - accuracy: 0.8550 - val_loss: 0.5994 - val_accuracy: 0.7917 Epoch 91/100 25/25 [==============================] - 33s 1s/step - loss: 0.3292 - accuracy: 0.8800 - val_loss: 0.5560 - val_accuracy: 0.8125 Epoch 92/100 25/25 [==============================] - 33s 1s/step - loss: 0.2759 - accuracy: 0.8913 - val_loss: 0.5550 - val_accuracy: 0.8229 Epoch 93/100 25/25 [==============================] - 35s 1s/step - loss: 0.3063 - accuracy: 0.8788 - val_loss: 0.6402 - val_accuracy: 0.8125 Epoch 94/100 25/25 [==============================] - 36s 1s/step - loss: 0.2841 - accuracy: 0.8863 - val_loss: 0.6440 - val_accuracy: 0.7760 Epoch 95/100 25/25 [==============================] - 37s 1s/step - loss: 0.2735 - accuracy: 0.8938 - val_loss: 0.6554 - val_accuracy: 0.8281 Epoch 96/100 25/25 [==============================] - 34s 1s/step - loss: 0.2625 - accuracy: 0.9025 - val_loss: 0.7760 - val_accuracy: 0.7240 Epoch 97/100 25/25 [==============================] - 34s 1s/step - loss: 0.2929 - accuracy: 0.8712 - val_loss: 0.5337 - val_accuracy: 0.8125 Epoch 98/100 25/25 [==============================] - 34s 1s/step - loss: 0.3274 - accuracy: 0.8838 - val_loss: 0.5662 - val_accuracy: 0.8021 Epoch 99/100 25/25 [==============================] - 34s 1s/step - loss: 0.2644 - accuracy: 0.8800 - val_loss: 0.5249 - val_accuracy: 0.8333 Epoch 100/100 25/25 [==============================] - 33s 1s/step - loss: 0.2609 - accuracy: 0.8950 - val_loss: 0.5722 - val_accuracy: 0.8281
<keras.callbacks.History at 0x2b047d192b0>
model_flat_drop.evaluate(test_ds)8/8 [==============================] - 8s 256ms/step - loss: 0.5354 - accuracy: 0.8008
[0.535420835018158, 0.80078125]
Do warstw maxpooling
model_pool_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(10, activation='softmax')
])model_pool_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_pool_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_5 (Conv2D) (None, 55, 55, 96) 34944
max_pooling2d_3 (MaxPooling (None, 27, 27, 96) 0
2D)
dropout_2 (Dropout) (None, 27, 27, 96) 0
conv2d_6 (Conv2D) (None, 27, 27, 256) 614656
max_pooling2d_4 (MaxPooling (None, 13, 13, 256) 0
2D)
dropout_3 (Dropout) (None, 13, 13, 256) 0
conv2d_7 (Conv2D) (None, 13, 13, 384) 885120
conv2d_8 (Conv2D) (None, 13, 13, 384) 1327488
conv2d_9 (Conv2D) (None, 13, 13, 256) 884992
max_pooling2d_5 (MaxPooling (None, 6, 6, 256) 0
2D)
dropout_4 (Dropout) (None, 6, 6, 256) 0
flatten_1 (Flatten) (None, 9216) 0
dense_3 (Dense) (None, 4096) 37752832
dense_4 (Dense) (None, 4096) 16781312
dense_5 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_pool_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 38s 1s/step - loss: 2.1540 - accuracy: 0.1675 - val_loss: 2.0134 - val_accuracy: 0.1979 Epoch 2/100 25/25 [==============================] - 37s 1s/step - loss: 1.6939 - accuracy: 0.2037 - val_loss: 1.7293 - val_accuracy: 0.1875 Epoch 3/100 25/25 [==============================] - 35s 1s/step - loss: 1.6265 - accuracy: 0.2512 - val_loss: 1.7483 - val_accuracy: 0.2031 Epoch 4/100 25/25 [==============================] - 34s 1s/step - loss: 1.6241 - accuracy: 0.2463 - val_loss: 1.7277 - val_accuracy: 0.2135 Epoch 5/100 25/25 [==============================] - 34s 1s/step - loss: 1.6075 - accuracy: 0.2675 - val_loss: 1.6551 - val_accuracy: 0.2292 Epoch 6/100 25/25 [==============================] - 34s 1s/step - loss: 1.5647 - accuracy: 0.3025 - val_loss: 1.6350 - val_accuracy: 0.4219 Epoch 7/100 25/25 [==============================] - 36s 1s/step - loss: 1.5461 - accuracy: 0.2937 - val_loss: 1.6199 - val_accuracy: 0.2812 Epoch 8/100 25/25 [==============================] - 36s 1s/step - loss: 1.5298 - accuracy: 0.3250 - val_loss: 1.6913 - val_accuracy: 0.3594 Epoch 9/100 25/25 [==============================] - 34s 1s/step - loss: 1.4970 - accuracy: 0.3338 - val_loss: 1.6024 - val_accuracy: 0.4167 Epoch 10/100 25/25 [==============================] - 34s 1s/step - loss: 1.4647 - accuracy: 0.3625 - val_loss: 1.6006 - val_accuracy: 0.3646 Epoch 11/100 25/25 [==============================] - 34s 1s/step - loss: 1.4492 - accuracy: 0.3925 - val_loss: 1.6130 - val_accuracy: 0.3542 Epoch 12/100 25/25 [==============================] - 34s 1s/step - loss: 1.3774 - accuracy: 0.4112 - val_loss: 1.6229 - val_accuracy: 0.3646 Epoch 13/100 25/25 [==============================] - 35s 1s/step - loss: 1.3320 - accuracy: 0.4375 - val_loss: 1.5338 - val_accuracy: 0.4219 Epoch 14/100 25/25 [==============================] - 37s 1s/step - loss: 1.3045 - accuracy: 0.4412 - val_loss: 1.4971 - val_accuracy: 0.3802 Epoch 15/100 25/25 [==============================] - 36s 1s/step - loss: 1.2122 - accuracy: 0.4938 - val_loss: 1.5767 - val_accuracy: 0.4271 Epoch 16/100 25/25 [==============================] - 34s 1s/step - loss: 1.2045 - accuracy: 0.5038 - val_loss: 1.3564 - val_accuracy: 0.4583 Epoch 17/100 25/25 [==============================] - 34s 1s/step - loss: 1.1888 - accuracy: 0.5050 - val_loss: 1.3598 - val_accuracy: 0.4323 Epoch 18/100 25/25 [==============================] - 32s 1s/step - loss: 1.1004 - accuracy: 0.5400 - val_loss: 1.3798 - val_accuracy: 0.4010 Epoch 19/100 25/25 [==============================] - 35s 1s/step - loss: 1.1161 - accuracy: 0.5138 - val_loss: 1.4139 - val_accuracy: 0.4688 Epoch 20/100 25/25 [==============================] - 36s 1s/step - loss: 1.1024 - accuracy: 0.5300 - val_loss: 1.3807 - val_accuracy: 0.4115 Epoch 21/100 25/25 [==============================] - 34s 1s/step - loss: 1.0852 - accuracy: 0.5350 - val_loss: 1.2784 - val_accuracy: 0.4688 Epoch 22/100 25/25 [==============================] - 34s 1s/step - loss: 0.9935 - accuracy: 0.5500 - val_loss: 1.0615 - val_accuracy: 0.5260 Epoch 23/100 25/25 [==============================] - 34s 1s/step - loss: 1.0719 - accuracy: 0.5300 - val_loss: 1.6332 - val_accuracy: 0.4479 Epoch 24/100 25/25 [==============================] - 34s 1s/step - loss: 0.9728 - accuracy: 0.5625 - val_loss: 1.3436 - val_accuracy: 0.4531 Epoch 25/100 25/25 [==============================] - 35s 1s/step - loss: 0.9514 - accuracy: 0.5788 - val_loss: 1.1052 - val_accuracy: 0.4792 Epoch 26/100 25/25 [==============================] - 36s 1s/step - loss: 1.0354 - accuracy: 0.5437 - val_loss: 1.2274 - val_accuracy: 0.4896 Epoch 27/100 25/25 [==============================] - 38s 1s/step - loss: 0.9764 - accuracy: 0.5675 - val_loss: 1.2700 - val_accuracy: 0.4531 Epoch 28/100 25/25 [==============================] - 35s 1s/step - loss: 0.9111 - accuracy: 0.5800 - val_loss: 1.3311 - val_accuracy: 0.4792 Epoch 29/100 25/25 [==============================] - 35s 1s/step - loss: 0.8978 - accuracy: 0.5987 - val_loss: 1.2087 - val_accuracy: 0.5208 Epoch 30/100 25/25 [==============================] - 34s 1s/step - loss: 0.9541 - accuracy: 0.5913 - val_loss: 1.0234 - val_accuracy: 0.5885 Epoch 31/100 25/25 [==============================] - 34s 1s/step - loss: 0.9083 - accuracy: 0.6000 - val_loss: 1.1497 - val_accuracy: 0.4844 Epoch 32/100 25/25 [==============================] - 34s 1s/step - loss: 0.8709 - accuracy: 0.6263 - val_loss: 0.9774 - val_accuracy: 0.6146 Epoch 33/100 25/25 [==============================] - 34s 1s/step - loss: 0.8831 - accuracy: 0.6400 - val_loss: 1.3298 - val_accuracy: 0.4635 Epoch 34/100 25/25 [==============================] - 37s 1s/step - loss: 0.9105 - accuracy: 0.6000 - val_loss: 1.0325 - val_accuracy: 0.5312 Epoch 35/100 25/25 [==============================] - 34s 1s/step - loss: 0.8981 - accuracy: 0.6225 - val_loss: 0.9792 - val_accuracy: 0.5833 Epoch 36/100 25/25 [==============================] - 34s 1s/step - loss: 0.8302 - accuracy: 0.6325 - val_loss: 1.0503 - val_accuracy: 0.5417 Epoch 37/100 25/25 [==============================] - 34s 1s/step - loss: 0.8196 - accuracy: 0.6450 - val_loss: 1.1518 - val_accuracy: 0.5208 Epoch 38/100 25/25 [==============================] - 34s 1s/step - loss: 0.7978 - accuracy: 0.6450 - val_loss: 1.0733 - val_accuracy: 0.5521 Epoch 39/100 25/25 [==============================] - 36s 1s/step - loss: 0.8764 - accuracy: 0.6200 - val_loss: 1.1687 - val_accuracy: 0.5521 Epoch 40/100 25/25 [==============================] - 35s 1s/step - loss: 0.8347 - accuracy: 0.6350 - val_loss: 0.9538 - val_accuracy: 0.5521 Epoch 41/100 25/25 [==============================] - 34s 1s/step - loss: 0.7740 - accuracy: 0.6600 - val_loss: 0.9828 - val_accuracy: 0.5573 Epoch 42/100 25/25 [==============================] - 34s 1s/step - loss: 0.7792 - accuracy: 0.6575 - val_loss: 0.9347 - val_accuracy: 0.6146 Epoch 43/100 25/25 [==============================] - 34s 1s/step - loss: 0.7643 - accuracy: 0.6637 - val_loss: 1.0073 - val_accuracy: 0.5521 Epoch 44/100 25/25 [==============================] - 34s 1s/step - loss: 0.8491 - accuracy: 0.6300 - val_loss: 0.9072 - val_accuracy: 0.5781 Epoch 45/100 25/25 [==============================] - 36s 1s/step - loss: 0.7689 - accuracy: 0.6662 - val_loss: 0.9700 - val_accuracy: 0.5885 Epoch 46/100 25/25 [==============================] - 35s 1s/step - loss: 0.7808 - accuracy: 0.6762 - val_loss: 0.8849 - val_accuracy: 0.5885 Epoch 47/100 25/25 [==============================] - 34s 1s/step - loss: 0.7912 - accuracy: 0.6700 - val_loss: 0.9794 - val_accuracy: 0.5938 Epoch 48/100 25/25 [==============================] - 35s 1s/step - loss: 0.7140 - accuracy: 0.6900 - val_loss: 1.0859 - val_accuracy: 0.5156 Epoch 49/100 25/25 [==============================] - 34s 1s/step - loss: 0.7231 - accuracy: 0.6812 - val_loss: 0.9919 - val_accuracy: 0.5312 Epoch 50/100 25/25 [==============================] - 34s 1s/step - loss: 0.7164 - accuracy: 0.6775 - val_loss: 0.8754 - val_accuracy: 0.5938 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 0.6902 - accuracy: 0.7000 - val_loss: 0.7496 - val_accuracy: 0.6667 Epoch 52/100 25/25 [==============================] - 35s 1s/step - loss: 0.6941 - accuracy: 0.6950 - val_loss: 0.8111 - val_accuracy: 0.6406 Epoch 53/100 25/25 [==============================] - 34s 1s/step - loss: 0.6511 - accuracy: 0.6963 - val_loss: 0.9502 - val_accuracy: 0.5365 Epoch 54/100 25/25 [==============================] - 34s 1s/step - loss: 0.7010 - accuracy: 0.6775 - val_loss: 1.1635 - val_accuracy: 0.5156 Epoch 55/100 25/25 [==============================] - 34s 1s/step - loss: 0.6365 - accuracy: 0.7063 - val_loss: 0.7768 - val_accuracy: 0.6615 Epoch 56/100 25/25 [==============================] - 34s 1s/step - loss: 0.6422 - accuracy: 0.7138 - val_loss: 0.8124 - val_accuracy: 0.6667 Epoch 57/100 25/25 [==============================] - 36s 1s/step - loss: 0.6389 - accuracy: 0.7050 - val_loss: 0.7729 - val_accuracy: 0.6719 Epoch 58/100 25/25 [==============================] - 34s 1s/step - loss: 0.6144 - accuracy: 0.7312 - val_loss: 1.0041 - val_accuracy: 0.5312 Epoch 59/100 25/25 [==============================] - 34s 1s/step - loss: 0.7239 - accuracy: 0.6963 - val_loss: 0.8224 - val_accuracy: 0.6510 Epoch 60/100 25/25 [==============================] - 34s 1s/step - loss: 0.6382 - accuracy: 0.7000 - val_loss: 0.6888 - val_accuracy: 0.6823 Epoch 61/100 25/25 [==============================] - 34s 1s/step - loss: 0.6108 - accuracy: 0.7225 - val_loss: 0.6762 - val_accuracy: 0.6823 Epoch 62/100 25/25 [==============================] - 37s 1s/step - loss: 0.5994 - accuracy: 0.7412 - val_loss: 0.8999 - val_accuracy: 0.5573 Epoch 63/100 25/25 [==============================] - 35s 1s/step - loss: 0.5431 - accuracy: 0.7487 - val_loss: 0.7129 - val_accuracy: 0.6771 Epoch 64/100 25/25 [==============================] - 35s 1s/step - loss: 0.5872 - accuracy: 0.7550 - val_loss: 0.7451 - val_accuracy: 0.6406 Epoch 65/100 25/25 [==============================] - 34s 1s/step - loss: 0.5637 - accuracy: 0.7425 - val_loss: 0.6809 - val_accuracy: 0.6927 Epoch 66/100 25/25 [==============================] - 34s 1s/step - loss: 0.5531 - accuracy: 0.7412 - val_loss: 0.8347 - val_accuracy: 0.6094 Epoch 67/100 25/25 [==============================] - 34s 1s/step - loss: 0.5204 - accuracy: 0.7625 - val_loss: 0.9630 - val_accuracy: 0.5833 Epoch 68/100 25/25 [==============================] - 36s 1s/step - loss: 0.5477 - accuracy: 0.7613 - val_loss: 0.7513 - val_accuracy: 0.6302 Epoch 69/100 25/25 [==============================] - 34s 1s/step - loss: 0.5896 - accuracy: 0.7500 - val_loss: 0.6534 - val_accuracy: 0.6927 Epoch 70/100 25/25 [==============================] - 34s 1s/step - loss: 0.5651 - accuracy: 0.7375 - val_loss: 0.6118 - val_accuracy: 0.7292 Epoch 71/100 25/25 [==============================] - 34s 1s/step - loss: 0.4896 - accuracy: 0.7788 - val_loss: 0.6155 - val_accuracy: 0.7292 Epoch 72/100 25/25 [==============================] - 34s 1s/step - loss: 0.4835 - accuracy: 0.7625 - val_loss: 0.8160 - val_accuracy: 0.6250 Epoch 73/100 25/25 [==============================] - 34s 1s/step - loss: 0.5748 - accuracy: 0.7412 - val_loss: 0.7547 - val_accuracy: 0.6302 Epoch 74/100 25/25 [==============================] - 34s 1s/step - loss: 0.4459 - accuracy: 0.7937 - val_loss: 0.7444 - val_accuracy: 0.6667 Epoch 75/100 25/25 [==============================] - 35s 1s/step - loss: 0.4576 - accuracy: 0.7775 - val_loss: 1.0260 - val_accuracy: 0.6667 Epoch 76/100 25/25 [==============================] - 35s 1s/step - loss: 0.8701 - accuracy: 0.6825 - val_loss: 0.8563 - val_accuracy: 0.6198 Epoch 77/100 25/25 [==============================] - 35s 1s/step - loss: 0.5644 - accuracy: 0.7462 - val_loss: 0.7395 - val_accuracy: 0.6406 Epoch 78/100 25/25 [==============================] - 35s 1s/step - loss: 0.4464 - accuracy: 0.7950 - val_loss: 0.7404 - val_accuracy: 0.6510 Epoch 79/100 25/25 [==============================] - 35s 1s/step - loss: 0.4584 - accuracy: 0.7862 - val_loss: 0.7534 - val_accuracy: 0.6510 Epoch 80/100 25/25 [==============================] - 35s 1s/step - loss: 0.5297 - accuracy: 0.7700 - val_loss: 0.6617 - val_accuracy: 0.7083 Epoch 81/100 25/25 [==============================] - 35s 1s/step - loss: 0.4441 - accuracy: 0.7950 - val_loss: 0.7048 - val_accuracy: 0.6927 Epoch 82/100 25/25 [==============================] - 35s 1s/step - loss: 0.5024 - accuracy: 0.7713 - val_loss: 0.7456 - val_accuracy: 0.6875 Epoch 83/100 25/25 [==============================] - 35s 1s/step - loss: 0.4858 - accuracy: 0.7750 - val_loss: 0.6363 - val_accuracy: 0.7552 Epoch 84/100 25/25 [==============================] - 35s 1s/step - loss: 0.4293 - accuracy: 0.8112 - val_loss: 0.6452 - val_accuracy: 0.6875 Epoch 85/100 25/25 [==============================] - 35s 1s/step - loss: 0.4369 - accuracy: 0.8000 - val_loss: 0.7804 - val_accuracy: 0.6510 Epoch 86/100 25/25 [==============================] - 35s 1s/step - loss: 0.3787 - accuracy: 0.8125 - val_loss: 0.7369 - val_accuracy: 0.6719 Epoch 87/100 25/25 [==============================] - 35s 1s/step - loss: 0.5366 - accuracy: 0.7837 - val_loss: 0.9294 - val_accuracy: 0.6615 Epoch 88/100 25/25 [==============================] - 35s 1s/step - loss: 0.4486 - accuracy: 0.8037 - val_loss: 0.6532 - val_accuracy: 0.6875 Epoch 89/100 25/25 [==============================] - 35s 1s/step - loss: 0.3971 - accuracy: 0.8263 - val_loss: 0.5793 - val_accuracy: 0.7188 Epoch 90/100 25/25 [==============================] - 34s 1s/step - loss: 0.4023 - accuracy: 0.8087 - val_loss: 0.6973 - val_accuracy: 0.7135 Epoch 91/100 25/25 [==============================] - 35s 1s/step - loss: 0.3739 - accuracy: 0.8338 - val_loss: 0.6377 - val_accuracy: 0.6927 Epoch 92/100 25/25 [==============================] - 35s 1s/step - loss: 0.4167 - accuracy: 0.7950 - val_loss: 0.6365 - val_accuracy: 0.7188 Epoch 93/100 25/25 [==============================] - 35s 1s/step - loss: 0.4039 - accuracy: 0.8163 - val_loss: 0.7140 - val_accuracy: 0.6719 Epoch 94/100 25/25 [==============================] - 35s 1s/step - loss: 0.3655 - accuracy: 0.8125 - val_loss: 0.5506 - val_accuracy: 0.7500 Epoch 95/100 25/25 [==============================] - 35s 1s/step - loss: 0.4764 - accuracy: 0.7925 - val_loss: 0.6725 - val_accuracy: 0.6927 Epoch 96/100 25/25 [==============================] - 35s 1s/step - loss: 0.3864 - accuracy: 0.8163 - val_loss: 0.7746 - val_accuracy: 0.6615 Epoch 97/100 25/25 [==============================] - 35s 1s/step - loss: 0.3479 - accuracy: 0.8413 - val_loss: 0.6701 - val_accuracy: 0.7083 Epoch 98/100 25/25 [==============================] - 34s 1s/step - loss: 0.3446 - accuracy: 0.8388 - val_loss: 0.5623 - val_accuracy: 0.7656 Epoch 99/100 25/25 [==============================] - 34s 1s/step - loss: 0.3953 - accuracy: 0.8150 - val_loss: 0.6013 - val_accuracy: 0.7448 Epoch 100/100 25/25 [==============================] - 35s 1s/step - loss: 0.3247 - accuracy: 0.8400 - val_loss: 0.6237 - val_accuracy: 0.7552
<keras.callbacks.History at 0x2b04806d160>
model_pool_drop.evaluate(test_ds)8/8 [==============================] - 3s 278ms/step - loss: 0.6054 - accuracy: 0.7578
[0.6054161787033081, 0.7578125]
Do warstw splotowych
model_conv_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(10, activation='softmax')
])model_conv_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_conv_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_2"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_10 (Conv2D) (None, 55, 55, 96) 34944
dropout_5 (Dropout) (None, 55, 55, 96) 0
max_pooling2d_6 (MaxPooling (None, 27, 27, 96) 0
2D)
conv2d_11 (Conv2D) (None, 27, 27, 256) 614656
dropout_6 (Dropout) (None, 27, 27, 256) 0
max_pooling2d_7 (MaxPooling (None, 13, 13, 256) 0
2D)
conv2d_12 (Conv2D) (None, 13, 13, 384) 885120
dropout_7 (Dropout) (None, 13, 13, 384) 0
conv2d_13 (Conv2D) (None, 13, 13, 384) 1327488
dropout_8 (Dropout) (None, 13, 13, 384) 0
conv2d_14 (Conv2D) (None, 13, 13, 256) 884992
dropout_9 (Dropout) (None, 13, 13, 256) 0
max_pooling2d_8 (MaxPooling (None, 6, 6, 256) 0
2D)
flatten_2 (Flatten) (None, 9216) 0
dense_6 (Dense) (None, 4096) 37752832
dense_7 (Dense) (None, 4096) 16781312
dense_8 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_conv_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 39s 1s/step - loss: 1.8422 - accuracy: 0.2313 - val_loss: 2.1335 - val_accuracy: 0.2604 Epoch 2/100 25/25 [==============================] - 37s 1s/step - loss: 1.7023 - accuracy: 0.2837 - val_loss: 2.0904 - val_accuracy: 0.2969 Epoch 3/100 25/25 [==============================] - 37s 1s/step - loss: 1.5376 - accuracy: 0.3313 - val_loss: 2.0778 - val_accuracy: 0.2292 Epoch 4/100 25/25 [==============================] - 37s 1s/step - loss: 1.4663 - accuracy: 0.3800 - val_loss: 2.0102 - val_accuracy: 0.3542 Epoch 5/100 25/25 [==============================] - 37s 1s/step - loss: 1.4467 - accuracy: 0.3938 - val_loss: 1.9956 - val_accuracy: 0.3073 Epoch 6/100 25/25 [==============================] - 38s 1s/step - loss: 1.2621 - accuracy: 0.4863 - val_loss: 1.8875 - val_accuracy: 0.3333 Epoch 7/100 25/25 [==============================] - 37s 1s/step - loss: 1.2946 - accuracy: 0.4675 - val_loss: 1.8695 - val_accuracy: 0.3958 Epoch 8/100 25/25 [==============================] - 38s 2s/step - loss: 1.1517 - accuracy: 0.5100 - val_loss: 1.7409 - val_accuracy: 0.4583 Epoch 9/100 25/25 [==============================] - 37s 1s/step - loss: 1.1045 - accuracy: 0.5350 - val_loss: 1.8332 - val_accuracy: 0.2031 Epoch 10/100 25/25 [==============================] - 37s 1s/step - loss: 1.0446 - accuracy: 0.5462 - val_loss: 1.7515 - val_accuracy: 0.3490 Epoch 11/100 25/25 [==============================] - 37s 1s/step - loss: 1.0663 - accuracy: 0.5575 - val_loss: 1.7029 - val_accuracy: 0.3594 Epoch 12/100 25/25 [==============================] - 37s 1s/step - loss: 1.0778 - accuracy: 0.5650 - val_loss: 1.7780 - val_accuracy: 0.3021 Epoch 13/100 25/25 [==============================] - 37s 1s/step - loss: 1.0175 - accuracy: 0.5663 - val_loss: 1.8585 - val_accuracy: 0.2760 Epoch 14/100 25/25 [==============================] - 37s 1s/step - loss: 0.9161 - accuracy: 0.6100 - val_loss: 1.6880 - val_accuracy: 0.3802 Epoch 15/100 25/25 [==============================] - 37s 1s/step - loss: 0.8277 - accuracy: 0.6488 - val_loss: 1.5378 - val_accuracy: 0.4323 Epoch 16/100 25/25 [==============================] - 37s 1s/step - loss: 0.8719 - accuracy: 0.6463 - val_loss: 1.6053 - val_accuracy: 0.5052 Epoch 17/100 25/25 [==============================] - 37s 1s/step - loss: 0.7539 - accuracy: 0.6812 - val_loss: 1.6414 - val_accuracy: 0.4115 Epoch 18/100 25/25 [==============================] - 37s 1s/step - loss: 0.7815 - accuracy: 0.6812 - val_loss: 1.4664 - val_accuracy: 0.6146 Epoch 19/100 25/25 [==============================] - 37s 1s/step - loss: 0.7458 - accuracy: 0.6913 - val_loss: 1.4077 - val_accuracy: 0.5677 Epoch 20/100 25/25 [==============================] - 37s 1s/step - loss: 0.9790 - accuracy: 0.5913 - val_loss: 1.7290 - val_accuracy: 0.2812 Epoch 21/100 25/25 [==============================] - 37s 1s/step - loss: 0.7419 - accuracy: 0.6950 - val_loss: 1.4896 - val_accuracy: 0.5000 Epoch 22/100 25/25 [==============================] - 37s 1s/step - loss: 0.6879 - accuracy: 0.7200 - val_loss: 1.3856 - val_accuracy: 0.5469 Epoch 23/100 25/25 [==============================] - 37s 1s/step - loss: 0.6642 - accuracy: 0.7125 - val_loss: 1.4391 - val_accuracy: 0.3594 Epoch 24/100 25/25 [==============================] - 37s 1s/step - loss: 0.6317 - accuracy: 0.7412 - val_loss: 1.3867 - val_accuracy: 0.5417 Epoch 25/100 25/25 [==============================] - 37s 1s/step - loss: 0.6106 - accuracy: 0.7462 - val_loss: 1.3900 - val_accuracy: 0.5469 Epoch 26/100 25/25 [==============================] - 37s 1s/step - loss: 0.6000 - accuracy: 0.7287 - val_loss: 1.3455 - val_accuracy: 0.5677 Epoch 27/100 25/25 [==============================] - 37s 1s/step - loss: 0.5725 - accuracy: 0.7900 - val_loss: 1.2634 - val_accuracy: 0.6667 Epoch 28/100 25/25 [==============================] - 37s 1s/step - loss: 0.5605 - accuracy: 0.7688 - val_loss: 1.2915 - val_accuracy: 0.6198 Epoch 29/100 25/25 [==============================] - 37s 1s/step - loss: 0.5432 - accuracy: 0.7875 - val_loss: 1.2972 - val_accuracy: 0.5469 Epoch 30/100 25/25 [==============================] - 37s 1s/step - loss: 0.5862 - accuracy: 0.7663 - val_loss: 1.3937 - val_accuracy: 0.4375 Epoch 31/100 25/25 [==============================] - 37s 1s/step - loss: 0.5134 - accuracy: 0.8000 - val_loss: 1.3887 - val_accuracy: 0.4792 Epoch 32/100 25/25 [==============================] - 37s 1s/step - loss: 0.5530 - accuracy: 0.7800 - val_loss: 1.3789 - val_accuracy: 0.4219 Epoch 33/100 25/25 [==============================] - 37s 1s/step - loss: 0.4936 - accuracy: 0.7763 - val_loss: 1.1190 - val_accuracy: 0.6771 Epoch 34/100 25/25 [==============================] - 38s 1s/step - loss: 0.5085 - accuracy: 0.7950 - val_loss: 1.3130 - val_accuracy: 0.5260 Epoch 35/100 25/25 [==============================] - 37s 1s/step - loss: 0.4900 - accuracy: 0.7962 - val_loss: 1.2185 - val_accuracy: 0.5573 Epoch 36/100 25/25 [==============================] - 37s 1s/step - loss: 0.4537 - accuracy: 0.8263 - val_loss: 1.5491 - val_accuracy: 0.3438 Epoch 37/100 25/25 [==============================] - 37s 1s/step - loss: 0.4313 - accuracy: 0.8325 - val_loss: 1.3085 - val_accuracy: 0.5052 Epoch 38/100 25/25 [==============================] - 37s 1s/step - loss: 0.4185 - accuracy: 0.8288 - val_loss: 1.1157 - val_accuracy: 0.6667 Epoch 39/100 25/25 [==============================] - 37s 1s/step - loss: 0.4420 - accuracy: 0.8037 - val_loss: 1.0747 - val_accuracy: 0.6719 Epoch 40/100 25/25 [==============================] - 37s 1s/step - loss: 0.6217 - accuracy: 0.7613 - val_loss: 1.2203 - val_accuracy: 0.5938 Epoch 41/100 25/25 [==============================] - 37s 1s/step - loss: 0.4495 - accuracy: 0.8125 - val_loss: 1.2375 - val_accuracy: 0.5573 Epoch 42/100 25/25 [==============================] - 37s 1s/step - loss: 0.3707 - accuracy: 0.8413 - val_loss: 1.1054 - val_accuracy: 0.5885 Epoch 43/100 25/25 [==============================] - 37s 1s/step - loss: 0.4187 - accuracy: 0.8138 - val_loss: 1.1526 - val_accuracy: 0.6198 Epoch 44/100 25/25 [==============================] - 37s 1s/step - loss: 0.3886 - accuracy: 0.8462 - val_loss: 1.0597 - val_accuracy: 0.6458 Epoch 45/100 25/25 [==============================] - 37s 1s/step - loss: 0.3577 - accuracy: 0.8425 - val_loss: 1.0726 - val_accuracy: 0.6302 Epoch 46/100 25/25 [==============================] - 37s 1s/step - loss: 0.4222 - accuracy: 0.8225 - val_loss: 1.1023 - val_accuracy: 0.6146 Epoch 47/100 25/25 [==============================] - 37s 1s/step - loss: 0.3652 - accuracy: 0.8462 - val_loss: 1.1679 - val_accuracy: 0.6250 Epoch 48/100 25/25 [==============================] - 37s 1s/step - loss: 0.3453 - accuracy: 0.8537 - val_loss: 1.1139 - val_accuracy: 0.5781 Epoch 49/100 25/25 [==============================] - 37s 1s/step - loss: 0.3015 - accuracy: 0.8763 - val_loss: 1.0735 - val_accuracy: 0.6302 Epoch 50/100 25/25 [==============================] - 37s 1s/step - loss: 0.3289 - accuracy: 0.8587 - val_loss: 1.0207 - val_accuracy: 0.6667 Epoch 51/100 25/25 [==============================] - 37s 1s/step - loss: 0.3748 - accuracy: 0.8475 - val_loss: 1.1093 - val_accuracy: 0.5885 Epoch 52/100 25/25 [==============================] - 37s 1s/step - loss: 0.4012 - accuracy: 0.8375 - val_loss: 1.2312 - val_accuracy: 0.5417 Epoch 53/100 25/25 [==============================] - 37s 1s/step - loss: 0.2989 - accuracy: 0.8813 - val_loss: 1.0820 - val_accuracy: 0.6615 Epoch 54/100 25/25 [==============================] - 38s 1s/step - loss: 0.2973 - accuracy: 0.8750 - val_loss: 0.9210 - val_accuracy: 0.7188 Epoch 55/100 25/25 [==============================] - 37s 1s/step - loss: 0.3219 - accuracy: 0.8650 - val_loss: 1.1205 - val_accuracy: 0.6198 Epoch 56/100 25/25 [==============================] - 37s 1s/step - loss: 0.3142 - accuracy: 0.8750 - val_loss: 0.9678 - val_accuracy: 0.6771 Epoch 57/100 25/25 [==============================] - 37s 1s/step - loss: 0.2701 - accuracy: 0.8788 - val_loss: 0.9047 - val_accuracy: 0.6927 Epoch 58/100 25/25 [==============================] - 38s 1s/step - loss: 0.2940 - accuracy: 0.8788 - val_loss: 1.0407 - val_accuracy: 0.6458 Epoch 59/100 25/25 [==============================] - 37s 1s/step - loss: 0.2552 - accuracy: 0.9025 - val_loss: 0.9503 - val_accuracy: 0.6719 Epoch 60/100 25/25 [==============================] - 37s 1s/step - loss: 0.2430 - accuracy: 0.8913 - val_loss: 0.9695 - val_accuracy: 0.6719 Epoch 61/100 25/25 [==============================] - 39s 2s/step - loss: 0.3017 - accuracy: 0.8850 - val_loss: 0.9939 - val_accuracy: 0.6771 Epoch 62/100 25/25 [==============================] - 37s 1s/step - loss: 0.2430 - accuracy: 0.8938 - val_loss: 0.8850 - val_accuracy: 0.7083 Epoch 63/100 25/25 [==============================] - 38s 1s/step - loss: 0.2560 - accuracy: 0.9025 - val_loss: 1.0165 - val_accuracy: 0.6146 Epoch 64/100 25/25 [==============================] - 37s 1s/step - loss: 0.2406 - accuracy: 0.8938 - val_loss: 0.9506 - val_accuracy: 0.6198 Epoch 65/100 25/25 [==============================] - 37s 1s/step - loss: 0.2375 - accuracy: 0.9125 - val_loss: 1.0983 - val_accuracy: 0.5312 Epoch 66/100 25/25 [==============================] - 37s 1s/step - loss: 0.3581 - accuracy: 0.8625 - val_loss: 1.0290 - val_accuracy: 0.6875 Epoch 67/100 25/25 [==============================] - 38s 1s/step - loss: 0.2074 - accuracy: 0.9175 - val_loss: 0.9548 - val_accuracy: 0.6146 Epoch 68/100 25/25 [==============================] - 38s 1s/step - loss: 0.2374 - accuracy: 0.9025 - val_loss: 1.0477 - val_accuracy: 0.6146 Epoch 69/100 25/25 [==============================] - 39s 2s/step - loss: 0.2269 - accuracy: 0.9075 - val_loss: 1.2301 - val_accuracy: 0.5312 Epoch 70/100 25/25 [==============================] - 41s 2s/step - loss: 0.1970 - accuracy: 0.9300 - val_loss: 1.2733 - val_accuracy: 0.5052 Epoch 71/100 25/25 [==============================] - 39s 2s/step - loss: 0.2358 - accuracy: 0.9112 - val_loss: 0.9070 - val_accuracy: 0.6979 Epoch 72/100 25/25 [==============================] - 37s 1s/step - loss: 0.2292 - accuracy: 0.9125 - val_loss: 1.0493 - val_accuracy: 0.5469 Epoch 73/100 25/25 [==============================] - 37s 1s/step - loss: 0.1826 - accuracy: 0.9250 - val_loss: 0.8005 - val_accuracy: 0.7552 Epoch 74/100 25/25 [==============================] - 37s 1s/step - loss: 0.2046 - accuracy: 0.9237 - val_loss: 0.9878 - val_accuracy: 0.6615 Epoch 75/100 25/25 [==============================] - 37s 1s/step - loss: 0.1801 - accuracy: 0.9225 - val_loss: 0.9359 - val_accuracy: 0.6719 Epoch 76/100 25/25 [==============================] - 37s 1s/step - loss: 0.2322 - accuracy: 0.9125 - val_loss: 0.8830 - val_accuracy: 0.6927 Epoch 77/100 25/25 [==============================] - 38s 2s/step - loss: 0.1762 - accuracy: 0.9325 - val_loss: 1.0842 - val_accuracy: 0.4844 Epoch 78/100 25/25 [==============================] - 38s 1s/step - loss: 0.1705 - accuracy: 0.9413 - val_loss: 0.9062 - val_accuracy: 0.6667 Epoch 79/100 25/25 [==============================] - 38s 2s/step - loss: 0.4088 - accuracy: 0.8625 - val_loss: 1.0009 - val_accuracy: 0.6875 Epoch 80/100 25/25 [==============================] - 38s 1s/step - loss: 0.1877 - accuracy: 0.9388 - val_loss: 0.9098 - val_accuracy: 0.6719 Epoch 81/100 25/25 [==============================] - 38s 1s/step - loss: 0.1855 - accuracy: 0.9362 - val_loss: 1.0068 - val_accuracy: 0.6302 Epoch 82/100 25/25 [==============================] - 37s 1s/step - loss: 0.1567 - accuracy: 0.9375 - val_loss: 0.9050 - val_accuracy: 0.6823 Epoch 83/100 25/25 [==============================] - 37s 1s/step - loss: 0.1537 - accuracy: 0.9475 - val_loss: 0.9882 - val_accuracy: 0.5885 Epoch 84/100 25/25 [==============================] - 37s 1s/step - loss: 0.1718 - accuracy: 0.9425 - val_loss: 1.1189 - val_accuracy: 0.5729 Epoch 85/100 25/25 [==============================] - 38s 1s/step - loss: 0.1222 - accuracy: 0.9575 - val_loss: 0.9589 - val_accuracy: 0.5677 Epoch 86/100 25/25 [==============================] - 38s 2s/step - loss: 0.2028 - accuracy: 0.9212 - val_loss: 1.0172 - val_accuracy: 0.6146 Epoch 87/100 25/25 [==============================] - 38s 1s/step - loss: 0.1412 - accuracy: 0.9488 - val_loss: 0.9860 - val_accuracy: 0.6458 Epoch 88/100 25/25 [==============================] - 38s 1s/step - loss: 0.1566 - accuracy: 0.9513 - val_loss: 0.9333 - val_accuracy: 0.6510 Epoch 89/100 25/25 [==============================] - 40s 2s/step - loss: 0.3019 - accuracy: 0.9062 - val_loss: 1.5026 - val_accuracy: 0.3854 Epoch 90/100 25/25 [==============================] - 38s 2s/step - loss: 0.1974 - accuracy: 0.9438 - val_loss: 0.9424 - val_accuracy: 0.6719 Epoch 91/100 25/25 [==============================] - 37s 1s/step - loss: 0.1236 - accuracy: 0.9550 - val_loss: 0.9276 - val_accuracy: 0.5990 Epoch 92/100 25/25 [==============================] - 37s 1s/step - loss: 0.1786 - accuracy: 0.9350 - val_loss: 0.7350 - val_accuracy: 0.7760 Epoch 93/100 25/25 [==============================] - 37s 1s/step - loss: 0.2246 - accuracy: 0.9287 - val_loss: 0.8939 - val_accuracy: 0.6771 Epoch 94/100 25/25 [==============================] - 37s 1s/step - loss: 0.1033 - accuracy: 0.9663 - val_loss: 0.8567 - val_accuracy: 0.6406 Epoch 95/100 25/25 [==============================] - 37s 1s/step - loss: 0.1007 - accuracy: 0.9688 - val_loss: 0.8316 - val_accuracy: 0.7083 Epoch 96/100 25/25 [==============================] - 37s 1s/step - loss: 0.1740 - accuracy: 0.9325 - val_loss: 0.8963 - val_accuracy: 0.6771 Epoch 97/100 25/25 [==============================] - 37s 1s/step - loss: 0.0915 - accuracy: 0.9700 - val_loss: 0.7647 - val_accuracy: 0.7552 Epoch 98/100 25/25 [==============================] - 37s 1s/step - loss: 0.1030 - accuracy: 0.9625 - val_loss: 0.8457 - val_accuracy: 0.7135 Epoch 99/100 25/25 [==============================] - 38s 1s/step - loss: 0.1826 - accuracy: 0.9312 - val_loss: 1.0152 - val_accuracy: 0.6719 Epoch 100/100 25/25 [==============================] - 37s 1s/step - loss: 0.1084 - accuracy: 0.9575 - val_loss: 1.0453 - val_accuracy: 0.5312
<keras.callbacks.History at 0x2b04825a3a0>
model_conv_drop.evaluate(test_ds)8/8 [==============================] - 3s 289ms/step - loss: 0.9870 - accuracy: 0.5664
[0.9869575500488281, 0.56640625]
Do warstw spłaszczonych i maxpooling
model_flat_pool_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(10, activation='softmax')
])model_flat_pool_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_flat_pool_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_15 (Conv2D) (None, 55, 55, 96) 34944
max_pooling2d_9 (MaxPooling (None, 27, 27, 96) 0
2D)
dropout_10 (Dropout) (None, 27, 27, 96) 0
conv2d_16 (Conv2D) (None, 27, 27, 256) 614656
max_pooling2d_10 (MaxPoolin (None, 13, 13, 256) 0
g2D)
dropout_11 (Dropout) (None, 13, 13, 256) 0
conv2d_17 (Conv2D) (None, 13, 13, 384) 885120
conv2d_18 (Conv2D) (None, 13, 13, 384) 1327488
conv2d_19 (Conv2D) (None, 13, 13, 256) 884992
max_pooling2d_11 (MaxPoolin (None, 6, 6, 256) 0
g2D)
dropout_12 (Dropout) (None, 6, 6, 256) 0
flatten_3 (Flatten) (None, 9216) 0
dense_9 (Dense) (None, 4096) 37752832
dropout_13 (Dropout) (None, 4096) 0
dense_10 (Dense) (None, 4096) 16781312
dropout_14 (Dropout) (None, 4096) 0
dense_11 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_flat_pool_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 37s 1s/step - loss: 2.0917 - accuracy: 0.1775 - val_loss: 1.9565 - val_accuracy: 0.2708 Epoch 2/100 25/25 [==============================] - 34s 1s/step - loss: 1.7322 - accuracy: 0.2000 - val_loss: 1.8304 - val_accuracy: 0.3177 Epoch 3/100 25/25 [==============================] - 34s 1s/step - loss: 1.6923 - accuracy: 0.2325 - val_loss: 1.8261 - val_accuracy: 0.3073 Epoch 4/100 25/25 [==============================] - 34s 1s/step - loss: 1.6810 - accuracy: 0.2175 - val_loss: 1.7829 - val_accuracy: 0.2656 Epoch 5/100 25/25 [==============================] - 34s 1s/step - loss: 1.6517 - accuracy: 0.2225 - val_loss: 1.7979 - val_accuracy: 0.2396 Epoch 6/100 25/25 [==============================] - 35s 1s/step - loss: 1.6633 - accuracy: 0.2225 - val_loss: 1.7933 - val_accuracy: 0.2240 Epoch 7/100 25/25 [==============================] - 35s 1s/step - loss: 1.6198 - accuracy: 0.2637 - val_loss: 1.7102 - val_accuracy: 0.3906 Epoch 8/100 25/25 [==============================] - 34s 1s/step - loss: 1.6002 - accuracy: 0.2763 - val_loss: 1.7037 - val_accuracy: 0.3490 Epoch 9/100 25/25 [==============================] - 34s 1s/step - loss: 1.6232 - accuracy: 0.2612 - val_loss: 1.7178 - val_accuracy: 0.4219 Epoch 10/100 25/25 [==============================] - 34s 1s/step - loss: 1.5658 - accuracy: 0.3038 - val_loss: 1.6378 - val_accuracy: 0.3333 Epoch 11/100 25/25 [==============================] - 35s 1s/step - loss: 1.5433 - accuracy: 0.3137 - val_loss: 1.6015 - val_accuracy: 0.3385 Epoch 12/100 25/25 [==============================] - 35s 1s/step - loss: 1.4841 - accuracy: 0.3363 - val_loss: 1.5526 - val_accuracy: 0.3438 Epoch 13/100 25/25 [==============================] - 35s 1s/step - loss: 1.4081 - accuracy: 0.3837 - val_loss: 1.4210 - val_accuracy: 0.3802 Epoch 14/100 25/25 [==============================] - 34s 1s/step - loss: 1.3291 - accuracy: 0.4375 - val_loss: 1.2532 - val_accuracy: 0.4948 Epoch 15/100 25/25 [==============================] - 34s 1s/step - loss: 1.2583 - accuracy: 0.4550 - val_loss: 1.2151 - val_accuracy: 0.4531 Epoch 16/100 25/25 [==============================] - 34s 1s/step - loss: 1.1925 - accuracy: 0.5088 - val_loss: 1.2792 - val_accuracy: 0.3698 Epoch 17/100 25/25 [==============================] - 34s 1s/step - loss: 1.1364 - accuracy: 0.5163 - val_loss: 1.2483 - val_accuracy: 0.4635 Epoch 18/100 25/25 [==============================] - 33s 1s/step - loss: 1.1357 - accuracy: 0.5038 - val_loss: 1.1645 - val_accuracy: 0.4479 Epoch 19/100 25/25 [==============================] - 33s 1s/step - loss: 1.1218 - accuracy: 0.4963 - val_loss: 1.2317 - val_accuracy: 0.4479 Epoch 20/100 25/25 [==============================] - 33s 1s/step - loss: 1.1589 - accuracy: 0.5038 - val_loss: 1.0579 - val_accuracy: 0.5312 Epoch 21/100 25/25 [==============================] - 33s 1s/step - loss: 1.0452 - accuracy: 0.5675 - val_loss: 1.0819 - val_accuracy: 0.5312 Epoch 22/100 25/25 [==============================] - 33s 1s/step - loss: 1.0479 - accuracy: 0.5688 - val_loss: 1.0547 - val_accuracy: 0.5573 Epoch 23/100 25/25 [==============================] - 33s 1s/step - loss: 1.0540 - accuracy: 0.5450 - val_loss: 0.9755 - val_accuracy: 0.5312 Epoch 24/100 25/25 [==============================] - 33s 1s/step - loss: 1.0686 - accuracy: 0.5638 - val_loss: 1.0784 - val_accuracy: 0.5573 Epoch 25/100 25/25 [==============================] - 33s 1s/step - loss: 0.9795 - accuracy: 0.5750 - val_loss: 0.9952 - val_accuracy: 0.5625 Epoch 26/100 25/25 [==============================] - 33s 1s/step - loss: 1.0000 - accuracy: 0.5638 - val_loss: 1.2539 - val_accuracy: 0.4635 Epoch 27/100 25/25 [==============================] - 33s 1s/step - loss: 1.0407 - accuracy: 0.5537 - val_loss: 1.2266 - val_accuracy: 0.4948 Epoch 28/100 25/25 [==============================] - 33s 1s/step - loss: 1.0390 - accuracy: 0.5450 - val_loss: 0.9923 - val_accuracy: 0.5833 Epoch 29/100 25/25 [==============================] - 33s 1s/step - loss: 0.9594 - accuracy: 0.5913 - val_loss: 1.0368 - val_accuracy: 0.5469 Epoch 30/100 25/25 [==============================] - 33s 1s/step - loss: 0.9869 - accuracy: 0.5738 - val_loss: 0.8615 - val_accuracy: 0.5573 Epoch 31/100 25/25 [==============================] - 33s 1s/step - loss: 0.9585 - accuracy: 0.5775 - val_loss: 1.0329 - val_accuracy: 0.5260 Epoch 32/100 25/25 [==============================] - 33s 1s/step - loss: 0.9576 - accuracy: 0.5713 - val_loss: 1.1186 - val_accuracy: 0.5208 Epoch 33/100 25/25 [==============================] - 33s 1s/step - loss: 0.9654 - accuracy: 0.5638 - val_loss: 0.8668 - val_accuracy: 0.5938 Epoch 34/100 25/25 [==============================] - 33s 1s/step - loss: 0.9641 - accuracy: 0.5600 - val_loss: 0.8702 - val_accuracy: 0.6094 Epoch 35/100 25/25 [==============================] - 33s 1s/step - loss: 0.9572 - accuracy: 0.5875 - val_loss: 0.9063 - val_accuracy: 0.5677 Epoch 36/100 25/25 [==============================] - 33s 1s/step - loss: 0.9522 - accuracy: 0.5962 - val_loss: 1.0374 - val_accuracy: 0.5521 Epoch 37/100 25/25 [==============================] - 33s 1s/step - loss: 0.9262 - accuracy: 0.6087 - val_loss: 1.5964 - val_accuracy: 0.3542 Epoch 38/100 25/25 [==============================] - 33s 1s/step - loss: 0.9213 - accuracy: 0.6150 - val_loss: 0.8814 - val_accuracy: 0.6250 Epoch 39/100 25/25 [==============================] - 33s 1s/step - loss: 0.9015 - accuracy: 0.6125 - val_loss: 1.1375 - val_accuracy: 0.4948 Epoch 40/100 25/25 [==============================] - 33s 1s/step - loss: 0.8985 - accuracy: 0.6100 - val_loss: 1.0702 - val_accuracy: 0.5469 Epoch 41/100 25/25 [==============================] - 33s 1s/step - loss: 0.9110 - accuracy: 0.5987 - val_loss: 1.1127 - val_accuracy: 0.5260 Epoch 42/100 25/25 [==============================] - 33s 1s/step - loss: 0.9490 - accuracy: 0.5888 - val_loss: 0.8745 - val_accuracy: 0.6354 Epoch 43/100 25/25 [==============================] - 33s 1s/step - loss: 0.8498 - accuracy: 0.6388 - val_loss: 0.9744 - val_accuracy: 0.6094 Epoch 44/100 25/25 [==============================] - 33s 1s/step - loss: 0.9338 - accuracy: 0.6000 - val_loss: 0.7914 - val_accuracy: 0.6771 Epoch 45/100 25/25 [==============================] - 33s 1s/step - loss: 0.8778 - accuracy: 0.6187 - val_loss: 0.9569 - val_accuracy: 0.6094 Epoch 46/100 25/25 [==============================] - 33s 1s/step - loss: 0.8683 - accuracy: 0.6488 - val_loss: 0.8768 - val_accuracy: 0.6302 Epoch 47/100 25/25 [==============================] - 33s 1s/step - loss: 0.9314 - accuracy: 0.5888 - val_loss: 0.8695 - val_accuracy: 0.6510 Epoch 48/100 25/25 [==============================] - 33s 1s/step - loss: 0.8562 - accuracy: 0.6087 - val_loss: 0.9378 - val_accuracy: 0.6146 Epoch 49/100 25/25 [==============================] - 34s 1s/step - loss: 0.8526 - accuracy: 0.6425 - val_loss: 0.7878 - val_accuracy: 0.6771 Epoch 50/100 25/25 [==============================] - 33s 1s/step - loss: 0.8686 - accuracy: 0.6075 - val_loss: 0.7986 - val_accuracy: 0.6823 Epoch 51/100 25/25 [==============================] - 33s 1s/step - loss: 0.8529 - accuracy: 0.6338 - val_loss: 0.9565 - val_accuracy: 0.6094 Epoch 52/100 25/25 [==============================] - 33s 1s/step - loss: 0.8452 - accuracy: 0.6425 - val_loss: 0.8586 - val_accuracy: 0.6510 Epoch 53/100 25/25 [==============================] - 33s 1s/step - loss: 0.7898 - accuracy: 0.6438 - val_loss: 0.7577 - val_accuracy: 0.6823 Epoch 54/100 25/25 [==============================] - 33s 1s/step - loss: 0.7980 - accuracy: 0.6488 - val_loss: 0.8313 - val_accuracy: 0.6510 Epoch 55/100 25/25 [==============================] - 33s 1s/step - loss: 0.8150 - accuracy: 0.6388 - val_loss: 0.7770 - val_accuracy: 0.6823 Epoch 56/100 25/25 [==============================] - 33s 1s/step - loss: 0.8610 - accuracy: 0.6538 - val_loss: 0.7361 - val_accuracy: 0.6927 Epoch 57/100 25/25 [==============================] - 33s 1s/step - loss: 0.8102 - accuracy: 0.6612 - val_loss: 0.9088 - val_accuracy: 0.6302 Epoch 58/100 25/25 [==============================] - 33s 1s/step - loss: 0.8354 - accuracy: 0.6388 - val_loss: 0.8243 - val_accuracy: 0.6250 Epoch 59/100 25/25 [==============================] - 33s 1s/step - loss: 0.7939 - accuracy: 0.6513 - val_loss: 0.7328 - val_accuracy: 0.6823 Epoch 60/100 25/25 [==============================] - 33s 1s/step - loss: 0.8086 - accuracy: 0.6338 - val_loss: 1.4440 - val_accuracy: 0.5417 Epoch 61/100 25/25 [==============================] - 33s 1s/step - loss: 0.8329 - accuracy: 0.6500 - val_loss: 1.1445 - val_accuracy: 0.5729 Epoch 62/100 25/25 [==============================] - 33s 1s/step - loss: 0.7804 - accuracy: 0.6425 - val_loss: 0.7927 - val_accuracy: 0.6615 Epoch 63/100 25/25 [==============================] - 34s 1s/step - loss: 0.7279 - accuracy: 0.6925 - val_loss: 0.8555 - val_accuracy: 0.6719 Epoch 64/100 25/25 [==============================] - 33s 1s/step - loss: 0.7491 - accuracy: 0.6862 - val_loss: 0.7689 - val_accuracy: 0.6823 Epoch 65/100 25/25 [==============================] - 33s 1s/step - loss: 0.7468 - accuracy: 0.6775 - val_loss: 0.8862 - val_accuracy: 0.5990 Epoch 66/100 25/25 [==============================] - 33s 1s/step - loss: 0.7560 - accuracy: 0.6950 - val_loss: 0.9006 - val_accuracy: 0.6198 Epoch 67/100 25/25 [==============================] - 33s 1s/step - loss: 0.6820 - accuracy: 0.6950 - val_loss: 0.9478 - val_accuracy: 0.5990 Epoch 68/100 25/25 [==============================] - 33s 1s/step - loss: 0.6749 - accuracy: 0.7250 - val_loss: 0.7632 - val_accuracy: 0.6719 Epoch 69/100 25/25 [==============================] - 33s 1s/step - loss: 0.8980 - accuracy: 0.6313 - val_loss: 1.2917 - val_accuracy: 0.5208 Epoch 70/100 25/25 [==============================] - 33s 1s/step - loss: 0.8385 - accuracy: 0.6250 - val_loss: 0.9292 - val_accuracy: 0.5677 Epoch 71/100 25/25 [==============================] - 33s 1s/step - loss: 0.8319 - accuracy: 0.6400 - val_loss: 0.9105 - val_accuracy: 0.5990 Epoch 72/100 25/25 [==============================] - 33s 1s/step - loss: 0.8015 - accuracy: 0.6562 - val_loss: 0.9724 - val_accuracy: 0.6250 Epoch 73/100 25/25 [==============================] - 33s 1s/step - loss: 0.7114 - accuracy: 0.7000 - val_loss: 0.7665 - val_accuracy: 0.6771 Epoch 74/100 25/25 [==============================] - 33s 1s/step - loss: 0.8950 - accuracy: 0.6150 - val_loss: 0.8339 - val_accuracy: 0.6042 Epoch 75/100 25/25 [==============================] - 33s 1s/step - loss: 0.8064 - accuracy: 0.6637 - val_loss: 0.7158 - val_accuracy: 0.7240 Epoch 76/100 25/25 [==============================] - 33s 1s/step - loss: 0.7647 - accuracy: 0.6800 - val_loss: 0.8539 - val_accuracy: 0.5677 Epoch 77/100 25/25 [==============================] - 34s 1s/step - loss: 0.7230 - accuracy: 0.6875 - val_loss: 0.9702 - val_accuracy: 0.5781 Epoch 78/100 25/25 [==============================] - 36s 1s/step - loss: 0.7361 - accuracy: 0.7063 - val_loss: 1.1083 - val_accuracy: 0.5677 Epoch 79/100 25/25 [==============================] - 35s 1s/step - loss: 0.7267 - accuracy: 0.7075 - val_loss: 0.8585 - val_accuracy: 0.6615 Epoch 80/100 25/25 [==============================] - 33s 1s/step - loss: 0.7779 - accuracy: 0.6775 - val_loss: 1.3162 - val_accuracy: 0.5104 Epoch 81/100 25/25 [==============================] - 33s 1s/step - loss: 0.7000 - accuracy: 0.6975 - val_loss: 0.8335 - val_accuracy: 0.6250 Epoch 82/100 25/25 [==============================] - 33s 1s/step - loss: 0.6793 - accuracy: 0.7262 - val_loss: 0.9848 - val_accuracy: 0.6146 Epoch 83/100 25/25 [==============================] - 32s 1s/step - loss: 0.6640 - accuracy: 0.7025 - val_loss: 0.7998 - val_accuracy: 0.6250 Epoch 84/100 25/25 [==============================] - 33s 1s/step - loss: 0.7114 - accuracy: 0.7063 - val_loss: 0.8843 - val_accuracy: 0.6146 Epoch 85/100 25/25 [==============================] - 33s 1s/step - loss: 0.7037 - accuracy: 0.7138 - val_loss: 0.7425 - val_accuracy: 0.6562 Epoch 86/100 25/25 [==============================] - 33s 1s/step - loss: 0.6398 - accuracy: 0.7437 - val_loss: 0.6782 - val_accuracy: 0.7240 Epoch 87/100 25/25 [==============================] - 33s 1s/step - loss: 0.6463 - accuracy: 0.7175 - val_loss: 1.0489 - val_accuracy: 0.5885 Epoch 88/100 25/25 [==============================] - 33s 1s/step - loss: 0.6026 - accuracy: 0.7462 - val_loss: 1.8062 - val_accuracy: 0.5365 Epoch 89/100 25/25 [==============================] - 33s 1s/step - loss: 0.5925 - accuracy: 0.7538 - val_loss: 1.4441 - val_accuracy: 0.5573 Epoch 90/100 25/25 [==============================] - 35s 1s/step - loss: 0.6420 - accuracy: 0.7262 - val_loss: 0.7644 - val_accuracy: 0.6719 Epoch 91/100 25/25 [==============================] - 34s 1s/step - loss: 0.5179 - accuracy: 0.7837 - val_loss: 0.6801 - val_accuracy: 0.6927 Epoch 92/100 25/25 [==============================] - 33s 1s/step - loss: 0.6311 - accuracy: 0.7387 - val_loss: 0.8016 - val_accuracy: 0.7031 Epoch 93/100 25/25 [==============================] - 33s 1s/step - loss: 0.6578 - accuracy: 0.7412 - val_loss: 0.9470 - val_accuracy: 0.6510 Epoch 94/100 25/25 [==============================] - 33s 1s/step - loss: 0.6059 - accuracy: 0.7450 - val_loss: 0.6827 - val_accuracy: 0.7083 Epoch 95/100 25/25 [==============================] - 33s 1s/step - loss: 0.5553 - accuracy: 0.7575 - val_loss: 0.7752 - val_accuracy: 0.6979 Epoch 96/100 25/25 [==============================] - 33s 1s/step - loss: 0.5328 - accuracy: 0.7812 - val_loss: 0.5755 - val_accuracy: 0.7448 Epoch 97/100 25/25 [==============================] - 33s 1s/step - loss: 0.4946 - accuracy: 0.7887 - val_loss: 0.8354 - val_accuracy: 0.6823 Epoch 98/100 25/25 [==============================] - 33s 1s/step - loss: 0.5000 - accuracy: 0.7812 - val_loss: 0.7233 - val_accuracy: 0.6875 Epoch 99/100 25/25 [==============================] - 33s 1s/step - loss: 0.5272 - accuracy: 0.7750 - val_loss: 1.0468 - val_accuracy: 0.5990 Epoch 100/100 25/25 [==============================] - 33s 1s/step - loss: 0.4910 - accuracy: 0.7900 - val_loss: 0.7302 - val_accuracy: 0.6562
<keras.callbacks.History at 0x2b04856fa60>
model_flat_pool_drop.evaluate(test_ds)8/8 [==============================] - 3s 272ms/step - loss: 0.6817 - accuracy: 0.7227
[0.6817080974578857, 0.72265625]
Do warstw spłaszczonych i splotowych
model_flat_conv_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(10, activation='softmax')
])model_flat_conv_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_flat_conv_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_4"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_20 (Conv2D) (None, 55, 55, 96) 34944
dropout_15 (Dropout) (None, 55, 55, 96) 0
max_pooling2d_12 (MaxPoolin (None, 27, 27, 96) 0
g2D)
conv2d_21 (Conv2D) (None, 27, 27, 256) 614656
dropout_16 (Dropout) (None, 27, 27, 256) 0
max_pooling2d_13 (MaxPoolin (None, 13, 13, 256) 0
g2D)
conv2d_22 (Conv2D) (None, 13, 13, 384) 885120
dropout_17 (Dropout) (None, 13, 13, 384) 0
conv2d_23 (Conv2D) (None, 13, 13, 384) 1327488
dropout_18 (Dropout) (None, 13, 13, 384) 0
conv2d_24 (Conv2D) (None, 13, 13, 256) 884992
dropout_19 (Dropout) (None, 13, 13, 256) 0
max_pooling2d_14 (MaxPoolin (None, 6, 6, 256) 0
g2D)
flatten_4 (Flatten) (None, 9216) 0
dense_12 (Dense) (None, 4096) 37752832
dropout_20 (Dropout) (None, 4096) 0
dense_13 (Dense) (None, 4096) 16781312
dropout_21 (Dropout) (None, 4096) 0
dense_14 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_flat_conv_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 38s 1s/step - loss: 1.8599 - accuracy: 0.2175 - val_loss: 2.1514 - val_accuracy: 0.2500 Epoch 2/100 25/25 [==============================] - 36s 1s/step - loss: 1.6948 - accuracy: 0.2500 - val_loss: 2.1303 - val_accuracy: 0.2656 Epoch 3/100 25/25 [==============================] - 36s 1s/step - loss: 1.5830 - accuracy: 0.3088 - val_loss: 2.0759 - val_accuracy: 0.1875 Epoch 4/100 25/25 [==============================] - 36s 1s/step - loss: 1.5122 - accuracy: 0.3425 - val_loss: 2.0394 - val_accuracy: 0.2344 Epoch 5/100 25/25 [==============================] - 36s 1s/step - loss: 1.4322 - accuracy: 0.3787 - val_loss: 1.9146 - val_accuracy: 0.3594 Epoch 6/100 25/25 [==============================] - 36s 1s/step - loss: 1.2522 - accuracy: 0.4450 - val_loss: 1.9610 - val_accuracy: 0.2448 Epoch 7/100 25/25 [==============================] - 35s 1s/step - loss: 1.2729 - accuracy: 0.4475 - val_loss: 1.7461 - val_accuracy: 0.4792 Epoch 8/100 25/25 [==============================] - 35s 1s/step - loss: 1.3159 - accuracy: 0.4475 - val_loss: 1.7940 - val_accuracy: 0.4323 Epoch 9/100 25/25 [==============================] - 36s 1s/step - loss: 1.1717 - accuracy: 0.4988 - val_loss: 1.6668 - val_accuracy: 0.5208 Epoch 10/100 25/25 [==============================] - 36s 1s/step - loss: 1.1578 - accuracy: 0.4825 - val_loss: 1.6895 - val_accuracy: 0.4062 Epoch 11/100 25/25 [==============================] - 35s 1s/step - loss: 1.0373 - accuracy: 0.5225 - val_loss: 1.6572 - val_accuracy: 0.4271 Epoch 12/100 25/25 [==============================] - 36s 1s/step - loss: 1.1024 - accuracy: 0.5050 - val_loss: 1.7203 - val_accuracy: 0.4062 Epoch 13/100 25/25 [==============================] - 35s 1s/step - loss: 1.0335 - accuracy: 0.5487 - val_loss: 1.5971 - val_accuracy: 0.4583 Epoch 14/100 25/25 [==============================] - 35s 1s/step - loss: 1.0407 - accuracy: 0.5238 - val_loss: 1.7055 - val_accuracy: 0.4010 Epoch 15/100 25/25 [==============================] - 36s 1s/step - loss: 1.0026 - accuracy: 0.5462 - val_loss: 1.5315 - val_accuracy: 0.4583 Epoch 16/100 25/25 [==============================] - 36s 1s/step - loss: 1.0267 - accuracy: 0.5350 - val_loss: 1.5610 - val_accuracy: 0.4844 Epoch 17/100 25/25 [==============================] - 35s 1s/step - loss: 0.9454 - accuracy: 0.5663 - val_loss: 1.7559 - val_accuracy: 0.2708 Epoch 18/100 25/25 [==============================] - 36s 1s/step - loss: 0.9948 - accuracy: 0.5675 - val_loss: 1.6200 - val_accuracy: 0.4219 Epoch 19/100 25/25 [==============================] - 36s 1s/step - loss: 0.9074 - accuracy: 0.5987 - val_loss: 1.7146 - val_accuracy: 0.2917 Epoch 20/100 25/25 [==============================] - 35s 1s/step - loss: 0.9107 - accuracy: 0.5950 - val_loss: 1.5316 - val_accuracy: 0.4427 Epoch 21/100 25/25 [==============================] - 36s 1s/step - loss: 0.8628 - accuracy: 0.6212 - val_loss: 1.6812 - val_accuracy: 0.3021 Epoch 22/100 25/25 [==============================] - 35s 1s/step - loss: 0.9339 - accuracy: 0.6200 - val_loss: 1.4395 - val_accuracy: 0.5104 Epoch 23/100 25/25 [==============================] - 36s 1s/step - loss: 0.8752 - accuracy: 0.6150 - val_loss: 1.5659 - val_accuracy: 0.4219 Epoch 24/100 25/25 [==============================] - 36s 1s/step - loss: 0.8968 - accuracy: 0.6200 - val_loss: 1.4972 - val_accuracy: 0.4531 Epoch 25/100 25/25 [==============================] - 36s 1s/step - loss: 0.8278 - accuracy: 0.6463 - val_loss: 1.5808 - val_accuracy: 0.3958 Epoch 26/100 25/25 [==============================] - 36s 1s/step - loss: 0.8339 - accuracy: 0.6450 - val_loss: 1.4826 - val_accuracy: 0.4427 Epoch 27/100 25/25 [==============================] - 36s 1s/step - loss: 0.7643 - accuracy: 0.6675 - val_loss: 1.3160 - val_accuracy: 0.5573 Epoch 28/100 25/25 [==============================] - 36s 1s/step - loss: 0.8465 - accuracy: 0.6325 - val_loss: 1.4294 - val_accuracy: 0.4844 Epoch 29/100 25/25 [==============================] - 36s 1s/step - loss: 0.7217 - accuracy: 0.7000 - val_loss: 1.4546 - val_accuracy: 0.4427 Epoch 30/100 25/25 [==============================] - 35s 1s/step - loss: 0.7759 - accuracy: 0.6700 - val_loss: 1.5743 - val_accuracy: 0.3854 Epoch 31/100 25/25 [==============================] - 35s 1s/step - loss: 0.7494 - accuracy: 0.6737 - val_loss: 1.4131 - val_accuracy: 0.4427 Epoch 32/100 25/25 [==============================] - 35s 1s/step - loss: 0.8117 - accuracy: 0.6712 - val_loss: 1.5993 - val_accuracy: 0.3229 Epoch 33/100 25/25 [==============================] - 36s 1s/step - loss: 0.6711 - accuracy: 0.7063 - val_loss: 1.4211 - val_accuracy: 0.4635 Epoch 34/100 25/25 [==============================] - 36s 1s/step - loss: 0.7168 - accuracy: 0.6988 - val_loss: 1.4401 - val_accuracy: 0.4115 Epoch 35/100 25/25 [==============================] - 36s 1s/step - loss: 0.6766 - accuracy: 0.7287 - val_loss: 1.3782 - val_accuracy: 0.4583 Epoch 36/100 25/25 [==============================] - 36s 1s/step - loss: 0.6349 - accuracy: 0.7312 - val_loss: 1.3980 - val_accuracy: 0.4219 Epoch 37/100 25/25 [==============================] - 35s 1s/step - loss: 0.6949 - accuracy: 0.7163 - val_loss: 1.4965 - val_accuracy: 0.3646 Epoch 38/100 25/25 [==============================] - 36s 1s/step - loss: 0.6801 - accuracy: 0.7375 - val_loss: 1.4329 - val_accuracy: 0.4219 Epoch 39/100 25/25 [==============================] - 36s 1s/step - loss: 0.6067 - accuracy: 0.7538 - val_loss: 1.2681 - val_accuracy: 0.5260 Epoch 40/100 25/25 [==============================] - 36s 1s/step - loss: 0.6001 - accuracy: 0.7450 - val_loss: 1.4195 - val_accuracy: 0.4062 Epoch 41/100 25/25 [==============================] - 36s 1s/step - loss: 0.6434 - accuracy: 0.7287 - val_loss: 1.4461 - val_accuracy: 0.3854 Epoch 42/100 25/25 [==============================] - 36s 1s/step - loss: 0.5218 - accuracy: 0.7738 - val_loss: 1.4192 - val_accuracy: 0.4010 Epoch 43/100 25/25 [==============================] - 35s 1s/step - loss: 0.5784 - accuracy: 0.7425 - val_loss: 1.5090 - val_accuracy: 0.3958 Epoch 44/100 25/25 [==============================] - 36s 1s/step - loss: 0.4925 - accuracy: 0.7950 - val_loss: 1.4046 - val_accuracy: 0.3750 Epoch 45/100 25/25 [==============================] - 35s 1s/step - loss: 0.5452 - accuracy: 0.7788 - val_loss: 1.4319 - val_accuracy: 0.4062 Epoch 46/100 25/25 [==============================] - 35s 1s/step - loss: 0.4841 - accuracy: 0.7725 - val_loss: 1.1874 - val_accuracy: 0.5729 Epoch 47/100 25/25 [==============================] - 36s 1s/step - loss: 0.4485 - accuracy: 0.8175 - val_loss: 1.1916 - val_accuracy: 0.5469 Epoch 48/100 25/25 [==============================] - 35s 1s/step - loss: 0.6050 - accuracy: 0.7513 - val_loss: 1.2601 - val_accuracy: 0.4948 Epoch 49/100 25/25 [==============================] - 36s 1s/step - loss: 0.5776 - accuracy: 0.7563 - val_loss: 1.2410 - val_accuracy: 0.5052 Epoch 50/100 25/25 [==============================] - 36s 1s/step - loss: 0.4671 - accuracy: 0.7925 - val_loss: 1.3867 - val_accuracy: 0.4219 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 0.4418 - accuracy: 0.8075 - val_loss: 1.1037 - val_accuracy: 0.6250 Epoch 52/100 25/25 [==============================] - 36s 1s/step - loss: 0.4319 - accuracy: 0.8175 - val_loss: 1.5872 - val_accuracy: 0.3385 Epoch 53/100 25/25 [==============================] - 35s 1s/step - loss: 0.4272 - accuracy: 0.8213 - val_loss: 1.3312 - val_accuracy: 0.4583 Epoch 54/100 25/25 [==============================] - 35s 1s/step - loss: 0.4158 - accuracy: 0.8275 - val_loss: 1.3446 - val_accuracy: 0.4167 Epoch 55/100 25/25 [==============================] - 35s 1s/step - loss: 0.5541 - accuracy: 0.8062 - val_loss: 1.3947 - val_accuracy: 0.4115 Epoch 56/100 25/25 [==============================] - 36s 1s/step - loss: 0.4572 - accuracy: 0.8012 - val_loss: 1.5154 - val_accuracy: 0.3698 Epoch 57/100 25/25 [==============================] - 36s 1s/step - loss: 0.4160 - accuracy: 0.8150 - val_loss: 1.2559 - val_accuracy: 0.4740 Epoch 58/100 25/25 [==============================] - 36s 1s/step - loss: 0.3743 - accuracy: 0.8313 - val_loss: 1.0860 - val_accuracy: 0.5729 Epoch 59/100 25/25 [==============================] - 36s 1s/step - loss: 0.4149 - accuracy: 0.8250 - val_loss: 1.2052 - val_accuracy: 0.5052 Epoch 60/100 25/25 [==============================] - 36s 1s/step - loss: 0.4006 - accuracy: 0.8062 - val_loss: 0.9254 - val_accuracy: 0.6615 Epoch 61/100 25/25 [==============================] - 36s 1s/step - loss: 0.3485 - accuracy: 0.8712 - val_loss: 1.2242 - val_accuracy: 0.5000 Epoch 62/100 25/25 [==============================] - 36s 1s/step - loss: 0.3578 - accuracy: 0.8612 - val_loss: 1.1742 - val_accuracy: 0.5312 Epoch 63/100 25/25 [==============================] - 35s 1s/step - loss: 0.9528 - accuracy: 0.7188 - val_loss: 1.4248 - val_accuracy: 0.4115 Epoch 64/100 25/25 [==============================] - 35s 1s/step - loss: 0.5106 - accuracy: 0.7962 - val_loss: 1.3560 - val_accuracy: 0.4062 Epoch 65/100 25/25 [==============================] - 35s 1s/step - loss: 0.4529 - accuracy: 0.8075 - val_loss: 1.2284 - val_accuracy: 0.4896 Epoch 66/100 25/25 [==============================] - 35s 1s/step - loss: 0.4449 - accuracy: 0.8238 - val_loss: 1.3032 - val_accuracy: 0.4271 Epoch 67/100 25/25 [==============================] - 37s 1s/step - loss: 0.3620 - accuracy: 0.8512 - val_loss: 1.3937 - val_accuracy: 0.3802 Epoch 68/100 25/25 [==============================] - 36s 1s/step - loss: 0.4211 - accuracy: 0.8413 - val_loss: 1.5752 - val_accuracy: 0.3229 Epoch 69/100 25/25 [==============================] - 35s 1s/step - loss: 0.3619 - accuracy: 0.8500 - val_loss: 1.3213 - val_accuracy: 0.4583 Epoch 70/100 25/25 [==============================] - 35s 1s/step - loss: 0.3436 - accuracy: 0.8587 - val_loss: 1.3979 - val_accuracy: 0.3854 Epoch 71/100 25/25 [==============================] - 35s 1s/step - loss: 0.3347 - accuracy: 0.8550 - val_loss: 1.3089 - val_accuracy: 0.4844 Epoch 72/100 25/25 [==============================] - 35s 1s/step - loss: 0.3278 - accuracy: 0.8562 - val_loss: 1.1956 - val_accuracy: 0.5052 Epoch 73/100 25/25 [==============================] - 35s 1s/step - loss: 0.3228 - accuracy: 0.8687 - val_loss: 1.2078 - val_accuracy: 0.5000 Epoch 74/100 25/25 [==============================] - 35s 1s/step - loss: 0.5337 - accuracy: 0.8037 - val_loss: 1.2706 - val_accuracy: 0.4792 Epoch 75/100 25/25 [==============================] - 35s 1s/step - loss: 0.3490 - accuracy: 0.8650 - val_loss: 1.4795 - val_accuracy: 0.4062 Epoch 76/100 25/25 [==============================] - 35s 1s/step - loss: 0.2937 - accuracy: 0.8737 - val_loss: 1.1618 - val_accuracy: 0.5417 Epoch 77/100 25/25 [==============================] - 35s 1s/step - loss: 0.5818 - accuracy: 0.8112 - val_loss: 1.3942 - val_accuracy: 0.4062 Epoch 78/100 25/25 [==============================] - 35s 1s/step - loss: 0.4328 - accuracy: 0.8225 - val_loss: 1.3289 - val_accuracy: 0.4271 Epoch 79/100 25/25 [==============================] - 37s 1s/step - loss: 0.3534 - accuracy: 0.8625 - val_loss: 1.3902 - val_accuracy: 0.3906 Epoch 80/100 25/25 [==============================] - 39s 2s/step - loss: 0.3293 - accuracy: 0.8725 - val_loss: 1.2934 - val_accuracy: 0.4740 Epoch 81/100 25/25 [==============================] - 36s 1s/step - loss: 0.3620 - accuracy: 0.8562 - val_loss: 1.2229 - val_accuracy: 0.4688 Epoch 82/100 25/25 [==============================] - 35s 1s/step - loss: 0.2888 - accuracy: 0.8813 - val_loss: 1.2308 - val_accuracy: 0.5000 Epoch 83/100 25/25 [==============================] - 35s 1s/step - loss: 0.2521 - accuracy: 0.8975 - val_loss: 1.0596 - val_accuracy: 0.5365 Epoch 84/100 25/25 [==============================] - 35s 1s/step - loss: 0.2750 - accuracy: 0.8725 - val_loss: 1.2119 - val_accuracy: 0.5052 Epoch 85/100 25/25 [==============================] - 35s 1s/step - loss: 0.2639 - accuracy: 0.9087 - val_loss: 1.2608 - val_accuracy: 0.4740 Epoch 86/100 25/25 [==============================] - 35s 1s/step - loss: 0.3473 - accuracy: 0.8512 - val_loss: 1.0966 - val_accuracy: 0.5990 Epoch 87/100 25/25 [==============================] - 35s 1s/step - loss: 0.2814 - accuracy: 0.8800 - val_loss: 1.0908 - val_accuracy: 0.5625 Epoch 88/100 25/25 [==============================] - 35s 1s/step - loss: 0.2442 - accuracy: 0.8988 - val_loss: 1.0714 - val_accuracy: 0.5521 Epoch 89/100 25/25 [==============================] - 35s 1s/step - loss: 0.2632 - accuracy: 0.9000 - val_loss: 1.2409 - val_accuracy: 0.5000 Epoch 90/100 25/25 [==============================] - 35s 1s/step - loss: 0.2928 - accuracy: 0.8687 - val_loss: 1.2651 - val_accuracy: 0.5208 Epoch 91/100 25/25 [==============================] - 35s 1s/step - loss: 0.2357 - accuracy: 0.9025 - val_loss: 1.1588 - val_accuracy: 0.5312 Epoch 92/100 25/25 [==============================] - 35s 1s/step - loss: 0.2177 - accuracy: 0.9075 - val_loss: 1.1720 - val_accuracy: 0.5469 Epoch 93/100 25/25 [==============================] - 35s 1s/step - loss: 0.7621 - accuracy: 0.7962 - val_loss: 1.3168 - val_accuracy: 0.5260 Epoch 94/100 25/25 [==============================] - 35s 1s/step - loss: 0.4913 - accuracy: 0.8150 - val_loss: 1.3253 - val_accuracy: 0.4479 Epoch 95/100 25/25 [==============================] - 35s 1s/step - loss: 0.3095 - accuracy: 0.8850 - val_loss: 1.0981 - val_accuracy: 0.5312 Epoch 96/100 25/25 [==============================] - 35s 1s/step - loss: 0.3121 - accuracy: 0.8662 - val_loss: 1.1353 - val_accuracy: 0.5260 Epoch 97/100 25/25 [==============================] - 35s 1s/step - loss: 0.2517 - accuracy: 0.9038 - val_loss: 1.1186 - val_accuracy: 0.5625 Epoch 98/100 25/25 [==============================] - 35s 1s/step - loss: 0.2424 - accuracy: 0.9112 - val_loss: 1.2868 - val_accuracy: 0.4844 Epoch 99/100 25/25 [==============================] - 35s 1s/step - loss: 0.2683 - accuracy: 0.8900 - val_loss: 1.1035 - val_accuracy: 0.5729 Epoch 100/100 25/25 [==============================] - 35s 1s/step - loss: 0.2120 - accuracy: 0.9075 - val_loss: 1.0550 - val_accuracy: 0.5781
<keras.callbacks.History at 0x2b049b321c0>
model_flat_conv_drop.evaluate(test_ds)8/8 [==============================] - 4s 260ms/step - loss: 0.9765 - accuracy: 0.6445
[0.9765039086341858, 0.64453125]
Do warstw maxpooling i splotowych
model_pool_conv_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(10, activation='softmax')
])model_pool_conv_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_pool_conv_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_5"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_25 (Conv2D) (None, 55, 55, 96) 34944
dropout_22 (Dropout) (None, 55, 55, 96) 0
max_pooling2d_15 (MaxPoolin (None, 27, 27, 96) 0
g2D)
dropout_23 (Dropout) (None, 27, 27, 96) 0
conv2d_26 (Conv2D) (None, 27, 27, 256) 614656
dropout_24 (Dropout) (None, 27, 27, 256) 0
max_pooling2d_16 (MaxPoolin (None, 13, 13, 256) 0
g2D)
dropout_25 (Dropout) (None, 13, 13, 256) 0
conv2d_27 (Conv2D) (None, 13, 13, 384) 885120
dropout_26 (Dropout) (None, 13, 13, 384) 0
conv2d_28 (Conv2D) (None, 13, 13, 384) 1327488
dropout_27 (Dropout) (None, 13, 13, 384) 0
conv2d_29 (Conv2D) (None, 13, 13, 256) 884992
dropout_28 (Dropout) (None, 13, 13, 256) 0
max_pooling2d_17 (MaxPoolin (None, 6, 6, 256) 0
g2D)
dropout_29 (Dropout) (None, 6, 6, 256) 0
flatten_5 (Flatten) (None, 9216) 0
dense_15 (Dense) (None, 4096) 37752832
dense_16 (Dense) (None, 4096) 16781312
dense_17 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_pool_conv_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 38s 1s/step - loss: 1.8169 - accuracy: 0.2000 - val_loss: 2.2493 - val_accuracy: 0.1875 Epoch 2/100 25/25 [==============================] - 36s 1s/step - loss: 1.6415 - accuracy: 0.2438 - val_loss: 2.2203 - val_accuracy: 0.1979 Epoch 3/100 25/25 [==============================] - 36s 1s/step - loss: 1.5604 - accuracy: 0.2862 - val_loss: 2.1751 - val_accuracy: 0.2083 Epoch 4/100 25/25 [==============================] - 36s 1s/step - loss: 1.5147 - accuracy: 0.3113 - val_loss: 2.1321 - val_accuracy: 0.1927 Epoch 5/100 25/25 [==============================] - 36s 1s/step - loss: 1.3579 - accuracy: 0.4162 - val_loss: 2.0497 - val_accuracy: 0.2865 Epoch 6/100 25/25 [==============================] - 36s 1s/step - loss: 1.3016 - accuracy: 0.4525 - val_loss: 2.0358 - val_accuracy: 0.2344 Epoch 7/100 25/25 [==============================] - 36s 1s/step - loss: 1.2612 - accuracy: 0.4512 - val_loss: 1.9886 - val_accuracy: 0.2292 Epoch 8/100 25/25 [==============================] - 36s 1s/step - loss: 1.3036 - accuracy: 0.4437 - val_loss: 2.0800 - val_accuracy: 0.1927 Epoch 9/100 25/25 [==============================] - 36s 1s/step - loss: 1.1765 - accuracy: 0.4850 - val_loss: 1.8733 - val_accuracy: 0.3021 Epoch 10/100 25/25 [==============================] - 36s 1s/step - loss: 1.2216 - accuracy: 0.5000 - val_loss: 2.0504 - val_accuracy: 0.1927 Epoch 11/100 25/25 [==============================] - 37s 1s/step - loss: 1.1682 - accuracy: 0.4850 - val_loss: 1.9643 - val_accuracy: 0.2396 Epoch 12/100 25/25 [==============================] - 36s 1s/step - loss: 1.1334 - accuracy: 0.5150 - val_loss: 1.9832 - val_accuracy: 0.2292 Epoch 13/100 25/25 [==============================] - 36s 1s/step - loss: 1.0921 - accuracy: 0.5113 - val_loss: 1.8860 - val_accuracy: 0.2500 Epoch 14/100 25/25 [==============================] - 36s 1s/step - loss: 1.0653 - accuracy: 0.5050 - val_loss: 1.8309 - val_accuracy: 0.2604 Epoch 15/100 25/25 [==============================] - 36s 1s/step - loss: 1.0711 - accuracy: 0.5325 - val_loss: 1.8706 - val_accuracy: 0.2604 Epoch 16/100 25/25 [==============================] - 36s 1s/step - loss: 1.0179 - accuracy: 0.5562 - val_loss: 1.8749 - val_accuracy: 0.2188 Epoch 17/100 25/25 [==============================] - 36s 1s/step - loss: 1.0475 - accuracy: 0.5462 - val_loss: 1.8350 - val_accuracy: 0.2240 Epoch 18/100 25/25 [==============================] - 36s 1s/step - loss: 1.0022 - accuracy: 0.5738 - val_loss: 1.6695 - val_accuracy: 0.4115 Epoch 19/100 25/25 [==============================] - 36s 1s/step - loss: 0.9710 - accuracy: 0.5938 - val_loss: 1.8079 - val_accuracy: 0.2240 Epoch 20/100 25/25 [==============================] - 36s 1s/step - loss: 1.0680 - accuracy: 0.5612 - val_loss: 1.7820 - val_accuracy: 0.2500 Epoch 21/100 25/25 [==============================] - 36s 1s/step - loss: 1.0123 - accuracy: 0.5713 - val_loss: 1.7886 - val_accuracy: 0.2500 Epoch 22/100 25/25 [==============================] - 36s 1s/step - loss: 0.9476 - accuracy: 0.6100 - val_loss: 1.6905 - val_accuracy: 0.3125 Epoch 23/100 25/25 [==============================] - 36s 1s/step - loss: 0.9235 - accuracy: 0.6000 - val_loss: 1.6969 - val_accuracy: 0.3073 Epoch 24/100 25/25 [==============================] - 36s 1s/step - loss: 0.8894 - accuracy: 0.5987 - val_loss: 1.7731 - val_accuracy: 0.2396 Epoch 25/100 25/25 [==============================] - 36s 1s/step - loss: 0.9201 - accuracy: 0.6212 - val_loss: 1.7130 - val_accuracy: 0.2448 Epoch 26/100 25/25 [==============================] - 36s 1s/step - loss: 0.9261 - accuracy: 0.6150 - val_loss: 1.7654 - val_accuracy: 0.2552 Epoch 27/100 25/25 [==============================] - 36s 1s/step - loss: 0.9241 - accuracy: 0.6250 - val_loss: 1.6630 - val_accuracy: 0.3021 Epoch 28/100 25/25 [==============================] - 36s 1s/step - loss: 0.9109 - accuracy: 0.6200 - val_loss: 1.5995 - val_accuracy: 0.3542 Epoch 29/100 25/25 [==============================] - 36s 1s/step - loss: 0.8485 - accuracy: 0.6450 - val_loss: 1.7325 - val_accuracy: 0.2344 Epoch 30/100 25/25 [==============================] - 36s 1s/step - loss: 0.8655 - accuracy: 0.6388 - val_loss: 1.6539 - val_accuracy: 0.3229 Epoch 31/100 25/25 [==============================] - 36s 1s/step - loss: 0.8572 - accuracy: 0.6338 - val_loss: 1.7899 - val_accuracy: 0.2240 Epoch 32/100 25/25 [==============================] - 36s 1s/step - loss: 0.9136 - accuracy: 0.6313 - val_loss: 1.7606 - val_accuracy: 0.2240 Epoch 33/100 25/25 [==============================] - 36s 1s/step - loss: 0.7934 - accuracy: 0.6550 - val_loss: 1.7149 - val_accuracy: 0.2292 Epoch 34/100 25/25 [==============================] - 36s 1s/step - loss: 0.8042 - accuracy: 0.6463 - val_loss: 1.7325 - val_accuracy: 0.2604 Epoch 35/100 25/25 [==============================] - 36s 1s/step - loss: 0.8510 - accuracy: 0.6388 - val_loss: 1.6244 - val_accuracy: 0.3021 Epoch 36/100 25/25 [==============================] - 36s 1s/step - loss: 0.7933 - accuracy: 0.6562 - val_loss: 1.7268 - val_accuracy: 0.2552 Epoch 37/100 25/25 [==============================] - 36s 1s/step - loss: 0.7170 - accuracy: 0.7075 - val_loss: 1.4832 - val_accuracy: 0.4323 Epoch 38/100 25/25 [==============================] - 36s 1s/step - loss: 0.8339 - accuracy: 0.6413 - val_loss: 1.6859 - val_accuracy: 0.2604 Epoch 39/100 25/25 [==============================] - 36s 1s/step - loss: 0.7030 - accuracy: 0.6825 - val_loss: 1.6517 - val_accuracy: 0.2500 Epoch 40/100 25/25 [==============================] - 36s 1s/step - loss: 0.7162 - accuracy: 0.6913 - val_loss: 1.6911 - val_accuracy: 0.2708 Epoch 41/100 25/25 [==============================] - 36s 1s/step - loss: 0.7770 - accuracy: 0.6650 - val_loss: 1.6254 - val_accuracy: 0.3125 Epoch 42/100 25/25 [==============================] - 36s 1s/step - loss: 0.6581 - accuracy: 0.7225 - val_loss: 1.6136 - val_accuracy: 0.3229 Epoch 43/100 25/25 [==============================] - 36s 1s/step - loss: 0.6846 - accuracy: 0.7100 - val_loss: 1.6485 - val_accuracy: 0.2865 Epoch 44/100 25/25 [==============================] - 36s 1s/step - loss: 0.6980 - accuracy: 0.6888 - val_loss: 1.7597 - val_accuracy: 0.2552 Epoch 45/100 25/25 [==============================] - 36s 1s/step - loss: 0.6496 - accuracy: 0.7400 - val_loss: 1.6483 - val_accuracy: 0.3073 Epoch 46/100 25/25 [==============================] - 36s 1s/step - loss: 0.6251 - accuracy: 0.7250 - val_loss: 1.6830 - val_accuracy: 0.2917 Epoch 47/100 25/25 [==============================] - 36s 1s/step - loss: 0.6244 - accuracy: 0.7163 - val_loss: 1.6909 - val_accuracy: 0.3177 Epoch 48/100 25/25 [==============================] - 36s 1s/step - loss: 0.7036 - accuracy: 0.7088 - val_loss: 1.6145 - val_accuracy: 0.2708 Epoch 49/100 25/25 [==============================] - 36s 1s/step - loss: 0.6414 - accuracy: 0.7088 - val_loss: 1.8017 - val_accuracy: 0.2500 Epoch 50/100 25/25 [==============================] - 36s 1s/step - loss: 0.6729 - accuracy: 0.6988 - val_loss: 1.6652 - val_accuracy: 0.3229 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 0.5843 - accuracy: 0.7350 - val_loss: 1.5871 - val_accuracy: 0.3385 Epoch 52/100 25/25 [==============================] - 36s 1s/step - loss: 0.6044 - accuracy: 0.7300 - val_loss: 1.6579 - val_accuracy: 0.2708 Epoch 53/100 25/25 [==============================] - 36s 1s/step - loss: 0.5451 - accuracy: 0.7475 - val_loss: 1.6316 - val_accuracy: 0.3125 Epoch 54/100 25/25 [==============================] - 36s 1s/step - loss: 0.5658 - accuracy: 0.7475 - val_loss: 1.4053 - val_accuracy: 0.4635 Epoch 55/100 25/25 [==============================] - 36s 1s/step - loss: 0.5437 - accuracy: 0.7500 - val_loss: 1.8277 - val_accuracy: 0.2448 Epoch 56/100 25/25 [==============================] - 36s 1s/step - loss: 0.5834 - accuracy: 0.7462 - val_loss: 1.9969 - val_accuracy: 0.2292 Epoch 57/100 25/25 [==============================] - 36s 1s/step - loss: 0.5382 - accuracy: 0.7462 - val_loss: 1.6513 - val_accuracy: 0.3229 Epoch 58/100 25/25 [==============================] - 36s 1s/step - loss: 0.5030 - accuracy: 0.7825 - val_loss: 1.6703 - val_accuracy: 0.3490 Epoch 59/100 25/25 [==============================] - 36s 1s/step - loss: 0.6065 - accuracy: 0.7275 - val_loss: 1.6761 - val_accuracy: 0.2708 Epoch 60/100 25/25 [==============================] - 36s 1s/step - loss: 0.5541 - accuracy: 0.7625 - val_loss: 1.6730 - val_accuracy: 0.2500 Epoch 61/100 25/25 [==============================] - 37s 1s/step - loss: 0.4906 - accuracy: 0.7713 - val_loss: 1.6541 - val_accuracy: 0.2865 Epoch 62/100 25/25 [==============================] - 36s 1s/step - loss: 0.5298 - accuracy: 0.7675 - val_loss: 1.7023 - val_accuracy: 0.2865 Epoch 63/100 25/25 [==============================] - 36s 1s/step - loss: 0.4985 - accuracy: 0.7875 - val_loss: 1.9830 - val_accuracy: 0.2344 Epoch 64/100 25/25 [==============================] - 36s 1s/step - loss: 0.4888 - accuracy: 0.7750 - val_loss: 1.6680 - val_accuracy: 0.2969 Epoch 65/100 25/25 [==============================] - 36s 1s/step - loss: 0.4419 - accuracy: 0.7875 - val_loss: 1.8855 - val_accuracy: 0.2656 Epoch 66/100 25/25 [==============================] - 36s 1s/step - loss: 0.4638 - accuracy: 0.7800 - val_loss: 1.6730 - val_accuracy: 0.3229 Epoch 67/100 25/25 [==============================] - 38s 2s/step - loss: 0.4665 - accuracy: 0.7875 - val_loss: 1.7077 - val_accuracy: 0.3229 Epoch 68/100 25/25 [==============================] - 36s 1s/step - loss: 0.4436 - accuracy: 0.7837 - val_loss: 1.8192 - val_accuracy: 0.3073 Epoch 69/100 25/25 [==============================] - 36s 1s/step - loss: 0.4956 - accuracy: 0.7800 - val_loss: 1.7902 - val_accuracy: 0.2917 Epoch 70/100 25/25 [==============================] - 36s 1s/step - loss: 0.4772 - accuracy: 0.7875 - val_loss: 1.7419 - val_accuracy: 0.2969 Epoch 71/100 25/25 [==============================] - 36s 1s/step - loss: 0.3955 - accuracy: 0.8263 - val_loss: 1.8084 - val_accuracy: 0.2708 Epoch 72/100 25/25 [==============================] - 36s 1s/step - loss: 0.4335 - accuracy: 0.8037 - val_loss: 1.8921 - val_accuracy: 0.2812 Epoch 73/100 25/25 [==============================] - 36s 1s/step - loss: 0.4031 - accuracy: 0.8138 - val_loss: 1.9873 - val_accuracy: 0.2656 Epoch 74/100 25/25 [==============================] - 38s 1s/step - loss: 1.2079 - accuracy: 0.6175 - val_loss: 1.9448 - val_accuracy: 0.2188 Epoch 75/100 25/25 [==============================] - 36s 1s/step - loss: 0.6683 - accuracy: 0.7350 - val_loss: 1.9602 - val_accuracy: 0.2396 Epoch 76/100 25/25 [==============================] - 36s 1s/step - loss: 0.5714 - accuracy: 0.7725 - val_loss: 1.5198 - val_accuracy: 0.3333 Epoch 77/100 25/25 [==============================] - 36s 1s/step - loss: 0.4513 - accuracy: 0.7950 - val_loss: 1.6600 - val_accuracy: 0.3021 Epoch 78/100 25/25 [==============================] - 36s 1s/step - loss: 0.4323 - accuracy: 0.8062 - val_loss: 1.7079 - val_accuracy: 0.3177 Epoch 79/100 25/25 [==============================] - 36s 1s/step - loss: 0.4245 - accuracy: 0.8037 - val_loss: 1.8053 - val_accuracy: 0.2708 Epoch 80/100 25/25 [==============================] - 37s 1s/step - loss: 0.4046 - accuracy: 0.8163 - val_loss: 1.8561 - val_accuracy: 0.2760 Epoch 81/100 25/25 [==============================] - 36s 1s/step - loss: 0.4789 - accuracy: 0.7775 - val_loss: 1.9273 - val_accuracy: 0.2552 Epoch 82/100 25/25 [==============================] - 36s 1s/step - loss: 0.4086 - accuracy: 0.8138 - val_loss: 2.1020 - val_accuracy: 0.2656 Epoch 83/100 25/25 [==============================] - 36s 1s/step - loss: 0.4059 - accuracy: 0.8238 - val_loss: 2.0258 - val_accuracy: 0.2917 Epoch 84/100 25/25 [==============================] - 36s 1s/step - loss: 0.3803 - accuracy: 0.8163 - val_loss: 1.7981 - val_accuracy: 0.2812 Epoch 85/100 25/25 [==============================] - 36s 1s/step - loss: 0.4042 - accuracy: 0.8138 - val_loss: 1.7933 - val_accuracy: 0.2083 Epoch 86/100 25/25 [==============================] - 37s 1s/step - loss: 0.4212 - accuracy: 0.8413 - val_loss: 1.6624 - val_accuracy: 0.2812 Epoch 87/100 25/25 [==============================] - 36s 1s/step - loss: 0.3808 - accuracy: 0.8363 - val_loss: 2.4115 - val_accuracy: 0.2292 Epoch 88/100 25/25 [==============================] - 36s 1s/step - loss: 0.6500 - accuracy: 0.7550 - val_loss: 1.9146 - val_accuracy: 0.2396 Epoch 89/100 25/25 [==============================] - 36s 1s/step - loss: 0.4128 - accuracy: 0.8188 - val_loss: 2.1177 - val_accuracy: 0.2500 Epoch 90/100 25/25 [==============================] - 36s 1s/step - loss: 0.3811 - accuracy: 0.8313 - val_loss: 2.1652 - val_accuracy: 0.2396 Epoch 91/100 25/25 [==============================] - 36s 1s/step - loss: 0.3584 - accuracy: 0.8363 - val_loss: 1.9992 - val_accuracy: 0.2604 Epoch 92/100 25/25 [==============================] - 37s 1s/step - loss: 0.3575 - accuracy: 0.8375 - val_loss: 1.9204 - val_accuracy: 0.2656 Epoch 93/100 25/25 [==============================] - 36s 1s/step - loss: 0.3357 - accuracy: 0.8413 - val_loss: 2.0910 - val_accuracy: 0.2396 Epoch 94/100 25/25 [==============================] - 36s 1s/step - loss: 0.4440 - accuracy: 0.8175 - val_loss: 2.1375 - val_accuracy: 0.2292 Epoch 95/100 25/25 [==============================] - 36s 1s/step - loss: 0.3838 - accuracy: 0.8300 - val_loss: 2.0104 - val_accuracy: 0.2708 Epoch 96/100 25/25 [==============================] - 36s 1s/step - loss: 0.3214 - accuracy: 0.8575 - val_loss: 2.2511 - val_accuracy: 0.2552 Epoch 97/100 25/25 [==============================] - 36s 1s/step - loss: 0.3630 - accuracy: 0.8300 - val_loss: 2.1232 - val_accuracy: 0.2604 Epoch 98/100 25/25 [==============================] - 38s 2s/step - loss: 0.3356 - accuracy: 0.8512 - val_loss: 2.0751 - val_accuracy: 0.2865 Epoch 99/100 25/25 [==============================] - 38s 2s/step - loss: 0.3300 - accuracy: 0.8475 - val_loss: 1.9835 - val_accuracy: 0.3073 Epoch 100/100 25/25 [==============================] - 37s 1s/step - loss: 0.3392 - accuracy: 0.8500 - val_loss: 2.2578 - val_accuracy: 0.2552
<keras.callbacks.History at 0x2b04c65e7f0>
model_pool_conv_drop.evaluate(test_ds)8/8 [==============================] - 2s 269ms/step - loss: 2.1567 - accuracy: 0.2891
[2.1567039489746094, 0.2890625]
Do warstw spłaszczonych, maxpooling i splotowych
model_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(10, activation='softmax')
])model_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_6"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_30 (Conv2D) (None, 55, 55, 96) 34944
dropout_30 (Dropout) (None, 55, 55, 96) 0
max_pooling2d_18 (MaxPoolin (None, 27, 27, 96) 0
g2D)
dropout_31 (Dropout) (None, 27, 27, 96) 0
conv2d_31 (Conv2D) (None, 27, 27, 256) 614656
dropout_32 (Dropout) (None, 27, 27, 256) 0
max_pooling2d_19 (MaxPoolin (None, 13, 13, 256) 0
g2D)
dropout_33 (Dropout) (None, 13, 13, 256) 0
conv2d_32 (Conv2D) (None, 13, 13, 384) 885120
dropout_34 (Dropout) (None, 13, 13, 384) 0
conv2d_33 (Conv2D) (None, 13, 13, 384) 1327488
dropout_35 (Dropout) (None, 13, 13, 384) 0
conv2d_34 (Conv2D) (None, 13, 13, 256) 884992
dropout_36 (Dropout) (None, 13, 13, 256) 0
max_pooling2d_20 (MaxPoolin (None, 6, 6, 256) 0
g2D)
dropout_37 (Dropout) (None, 6, 6, 256) 0
flatten_6 (Flatten) (None, 9216) 0
dense_18 (Dense) (None, 4096) 37752832
dropout_38 (Dropout) (None, 4096) 0
dense_19 (Dense) (None, 4096) 16781312
dropout_39 (Dropout) (None, 4096) 0
dense_20 (Dense) (None, 10) 40970
=================================================================
Total params: 58,322,314
Trainable params: 58,322,314
Non-trainable params: 0
_________________________________________________________________
model_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 39s 1s/step - loss: 1.9247 - accuracy: 0.1900 - val_loss: 2.2491 - val_accuracy: 0.1875 Epoch 2/100 25/25 [==============================] - 38s 2s/step - loss: 1.7130 - accuracy: 0.2062 - val_loss: 2.2165 - val_accuracy: 0.1771 Epoch 3/100 25/25 [==============================] - 39s 2s/step - loss: 1.6761 - accuracy: 0.2350 - val_loss: 2.1936 - val_accuracy: 0.2292 Epoch 4/100 25/25 [==============================] - 37s 1s/step - loss: 1.6171 - accuracy: 0.2500 - val_loss: 2.1467 - val_accuracy: 0.1927 Epoch 5/100 25/25 [==============================] - 37s 1s/step - loss: 1.5317 - accuracy: 0.3175 - val_loss: 2.0777 - val_accuracy: 0.1927 Epoch 6/100 25/25 [==============================] - 37s 1s/step - loss: 1.4446 - accuracy: 0.3938 - val_loss: 2.0430 - val_accuracy: 0.1979 Epoch 7/100 25/25 [==============================] - 37s 1s/step - loss: 1.3851 - accuracy: 0.3787 - val_loss: 1.9749 - val_accuracy: 0.2188 Epoch 8/100 25/25 [==============================] - 36s 1s/step - loss: 1.3739 - accuracy: 0.4100 - val_loss: 1.9833 - val_accuracy: 0.2135 Epoch 9/100 25/25 [==============================] - 36s 1s/step - loss: 1.3105 - accuracy: 0.4212 - val_loss: 1.9121 - val_accuracy: 0.2448 Epoch 10/100 25/25 [==============================] - 37s 1s/step - loss: 1.2919 - accuracy: 0.4212 - val_loss: 1.9497 - val_accuracy: 0.2240 Epoch 11/100 25/25 [==============================] - 36s 1s/step - loss: 1.2476 - accuracy: 0.4750 - val_loss: 1.9766 - val_accuracy: 0.2083 Epoch 12/100 25/25 [==============================] - 36s 1s/step - loss: 1.2704 - accuracy: 0.4325 - val_loss: 2.0245 - val_accuracy: 0.1823 Epoch 13/100 25/25 [==============================] - 36s 1s/step - loss: 1.1780 - accuracy: 0.4725 - val_loss: 1.9085 - val_accuracy: 0.2240 Epoch 14/100 25/25 [==============================] - 36s 1s/step - loss: 1.1744 - accuracy: 0.4750 - val_loss: 1.8902 - val_accuracy: 0.2135 Epoch 15/100 25/25 [==============================] - 36s 1s/step - loss: 1.2046 - accuracy: 0.4638 - val_loss: 1.9528 - val_accuracy: 0.2031 Epoch 16/100 25/25 [==============================] - 36s 1s/step - loss: 1.2021 - accuracy: 0.5000 - val_loss: 1.9438 - val_accuracy: 0.2031 Epoch 17/100 25/25 [==============================] - 37s 1s/step - loss: 1.1076 - accuracy: 0.5113 - val_loss: 1.9135 - val_accuracy: 0.2031 Epoch 18/100 25/25 [==============================] - 40s 2s/step - loss: 1.0509 - accuracy: 0.5250 - val_loss: 1.8978 - val_accuracy: 0.2135 Epoch 19/100 25/25 [==============================] - 37s 1s/step - loss: 1.1375 - accuracy: 0.5163 - val_loss: 1.8838 - val_accuracy: 0.2135 Epoch 20/100 25/25 [==============================] - 36s 1s/step - loss: 1.0640 - accuracy: 0.5462 - val_loss: 1.9466 - val_accuracy: 0.2031 Epoch 21/100 25/25 [==============================] - 36s 1s/step - loss: 1.0863 - accuracy: 0.5163 - val_loss: 1.9305 - val_accuracy: 0.2083 Epoch 22/100 25/25 [==============================] - 36s 1s/step - loss: 1.0853 - accuracy: 0.5288 - val_loss: 1.9383 - val_accuracy: 0.2083 Epoch 23/100 25/25 [==============================] - 36s 1s/step - loss: 1.0640 - accuracy: 0.5288 - val_loss: 1.9474 - val_accuracy: 0.2031 Epoch 24/100 25/25 [==============================] - 36s 1s/step - loss: 1.0696 - accuracy: 0.5437 - val_loss: 1.8716 - val_accuracy: 0.2031 Epoch 25/100 25/25 [==============================] - 36s 1s/step - loss: 1.0901 - accuracy: 0.5387 - val_loss: 1.9141 - val_accuracy: 0.1979 Epoch 26/100 25/25 [==============================] - 36s 1s/step - loss: 1.0558 - accuracy: 0.5312 - val_loss: 1.8808 - val_accuracy: 0.1927 Epoch 27/100 25/25 [==============================] - 36s 1s/step - loss: 1.0350 - accuracy: 0.5562 - val_loss: 1.9078 - val_accuracy: 0.2083 Epoch 28/100 25/25 [==============================] - 36s 1s/step - loss: 1.0032 - accuracy: 0.5487 - val_loss: 1.9923 - val_accuracy: 0.1823 Epoch 29/100 25/25 [==============================] - 36s 1s/step - loss: 1.0444 - accuracy: 0.5525 - val_loss: 1.8429 - val_accuracy: 0.2135 Epoch 30/100 25/25 [==============================] - 36s 1s/step - loss: 1.0202 - accuracy: 0.5512 - val_loss: 1.8479 - val_accuracy: 0.2135 Epoch 31/100 25/25 [==============================] - 36s 1s/step - loss: 1.0019 - accuracy: 0.5537 - val_loss: 1.9275 - val_accuracy: 0.2135 Epoch 32/100 25/25 [==============================] - 36s 1s/step - loss: 1.0272 - accuracy: 0.5437 - val_loss: 1.9026 - val_accuracy: 0.2031 Epoch 33/100 25/25 [==============================] - 36s 1s/step - loss: 0.9633 - accuracy: 0.5663 - val_loss: 1.9349 - val_accuracy: 0.2031 Epoch 34/100 25/25 [==============================] - 36s 1s/step - loss: 1.0038 - accuracy: 0.5688 - val_loss: 1.9385 - val_accuracy: 0.1979 Epoch 35/100 25/25 [==============================] - 36s 1s/step - loss: 0.9373 - accuracy: 0.5900 - val_loss: 2.0208 - val_accuracy: 0.2031 Epoch 36/100 25/25 [==============================] - 36s 1s/step - loss: 0.9253 - accuracy: 0.5888 - val_loss: 1.9166 - val_accuracy: 0.2031 Epoch 37/100 25/25 [==============================] - 36s 1s/step - loss: 0.9618 - accuracy: 0.5788 - val_loss: 1.8177 - val_accuracy: 0.1979 Epoch 38/100 25/25 [==============================] - 36s 1s/step - loss: 0.9833 - accuracy: 0.5688 - val_loss: 1.9429 - val_accuracy: 0.1979 Epoch 39/100 25/25 [==============================] - 36s 1s/step - loss: 0.9389 - accuracy: 0.5888 - val_loss: 1.8995 - val_accuracy: 0.1979 Epoch 40/100 25/25 [==============================] - 36s 1s/step - loss: 0.9457 - accuracy: 0.5775 - val_loss: 1.9466 - val_accuracy: 0.2083 Epoch 41/100 25/25 [==============================] - 36s 1s/step - loss: 0.9740 - accuracy: 0.5738 - val_loss: 1.7971 - val_accuracy: 0.2240 Epoch 42/100 25/25 [==============================] - 36s 1s/step - loss: 0.9194 - accuracy: 0.5975 - val_loss: 2.0969 - val_accuracy: 0.1823 Epoch 43/100 25/25 [==============================] - 36s 1s/step - loss: 0.9476 - accuracy: 0.5900 - val_loss: 1.8180 - val_accuracy: 0.2083 Epoch 44/100 25/25 [==============================] - 36s 1s/step - loss: 0.9245 - accuracy: 0.6000 - val_loss: 1.8373 - val_accuracy: 0.2188 Epoch 45/100 25/25 [==============================] - 36s 1s/step - loss: 0.9001 - accuracy: 0.5925 - val_loss: 1.8846 - val_accuracy: 0.2083 Epoch 46/100 25/25 [==============================] - 36s 1s/step - loss: 0.8989 - accuracy: 0.5962 - val_loss: 2.0002 - val_accuracy: 0.1979 Epoch 47/100 25/25 [==============================] - 36s 1s/step - loss: 0.8539 - accuracy: 0.6288 - val_loss: 1.9201 - val_accuracy: 0.1927 Epoch 48/100 25/25 [==============================] - 36s 1s/step - loss: 0.9051 - accuracy: 0.6263 - val_loss: 1.9303 - val_accuracy: 0.2083 Epoch 49/100 25/25 [==============================] - 36s 1s/step - loss: 0.8572 - accuracy: 0.6363 - val_loss: 1.8043 - val_accuracy: 0.2135 Epoch 50/100 25/25 [==============================] - 36s 1s/step - loss: 0.8245 - accuracy: 0.6513 - val_loss: 1.9120 - val_accuracy: 0.2083 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 0.8384 - accuracy: 0.6463 - val_loss: 1.8960 - val_accuracy: 0.2031 Epoch 52/100 25/25 [==============================] - 36s 1s/step - loss: 0.8193 - accuracy: 0.6375 - val_loss: 1.6472 - val_accuracy: 0.2969 Epoch 53/100 25/25 [==============================] - 36s 1s/step - loss: 0.8076 - accuracy: 0.6712 - val_loss: 2.0521 - val_accuracy: 0.2031 Epoch 54/100 25/25 [==============================] - 38s 1s/step - loss: 0.8572 - accuracy: 0.6225 - val_loss: 2.0045 - val_accuracy: 0.2135 Epoch 55/100 25/25 [==============================] - 36s 1s/step - loss: 0.8142 - accuracy: 0.6600 - val_loss: 1.9518 - val_accuracy: 0.2083 Epoch 56/100 25/25 [==============================] - 36s 1s/step - loss: 0.8600 - accuracy: 0.6538 - val_loss: 1.7073 - val_accuracy: 0.2604 Epoch 57/100 25/25 [==============================] - 36s 1s/step - loss: 0.7840 - accuracy: 0.6850 - val_loss: 1.9704 - val_accuracy: 0.2031 Epoch 58/100 25/25 [==============================] - 36s 1s/step - loss: 0.7923 - accuracy: 0.6825 - val_loss: 1.8118 - val_accuracy: 0.2500 Epoch 59/100 25/25 [==============================] - 36s 1s/step - loss: 0.7204 - accuracy: 0.6938 - val_loss: 1.9559 - val_accuracy: 0.2292 Epoch 60/100 25/25 [==============================] - 37s 1s/step - loss: 0.7991 - accuracy: 0.6488 - val_loss: 1.9263 - val_accuracy: 0.2135 Epoch 61/100 25/25 [==============================] - 36s 1s/step - loss: 0.7813 - accuracy: 0.6725 - val_loss: 1.8279 - val_accuracy: 0.2448 Epoch 62/100 25/25 [==============================] - 36s 1s/step - loss: 0.7738 - accuracy: 0.6750 - val_loss: 2.1088 - val_accuracy: 0.2188 Epoch 63/100 25/25 [==============================] - 36s 1s/step - loss: 0.7300 - accuracy: 0.6938 - val_loss: 2.0727 - val_accuracy: 0.2135 Epoch 64/100 25/25 [==============================] - 36s 1s/step - loss: 0.7127 - accuracy: 0.7025 - val_loss: 1.9929 - val_accuracy: 0.2292 Epoch 65/100 25/25 [==============================] - 38s 1s/step - loss: 0.7034 - accuracy: 0.7200 - val_loss: 2.1949 - val_accuracy: 0.1979 Epoch 66/100 25/25 [==============================] - 38s 1s/step - loss: 0.7238 - accuracy: 0.6888 - val_loss: 2.1694 - val_accuracy: 0.2240 Epoch 67/100 25/25 [==============================] - 36s 1s/step - loss: 0.7244 - accuracy: 0.7000 - val_loss: 2.2779 - val_accuracy: 0.2031 Epoch 68/100 25/25 [==============================] - 36s 1s/step - loss: 0.6549 - accuracy: 0.7237 - val_loss: 2.1810 - val_accuracy: 0.2240 Epoch 69/100 25/25 [==============================] - 36s 1s/step - loss: 0.5940 - accuracy: 0.7487 - val_loss: 1.9802 - val_accuracy: 0.2552 Epoch 70/100 25/25 [==============================] - 36s 1s/step - loss: 0.6346 - accuracy: 0.7300 - val_loss: 2.4555 - val_accuracy: 0.2083 Epoch 71/100 25/25 [==============================] - 36s 1s/step - loss: 0.5913 - accuracy: 0.7437 - val_loss: 2.3821 - val_accuracy: 0.2240 Epoch 72/100 25/25 [==============================] - 36s 1s/step - loss: 0.5870 - accuracy: 0.7312 - val_loss: 2.0892 - val_accuracy: 0.2292 Epoch 73/100 25/25 [==============================] - 36s 1s/step - loss: 0.8009 - accuracy: 0.6600 - val_loss: 2.1879 - val_accuracy: 0.2292 Epoch 74/100 25/25 [==============================] - 36s 1s/step - loss: 0.6669 - accuracy: 0.7125 - val_loss: 2.1540 - val_accuracy: 0.2344 Epoch 75/100 25/25 [==============================] - 36s 1s/step - loss: 0.6351 - accuracy: 0.7325 - val_loss: 1.9588 - val_accuracy: 0.2344 Epoch 76/100 25/25 [==============================] - 38s 2s/step - loss: 0.6817 - accuracy: 0.7237 - val_loss: 2.0462 - val_accuracy: 0.2292 Epoch 77/100 25/25 [==============================] - 36s 1s/step - loss: 0.6122 - accuracy: 0.7538 - val_loss: 2.0432 - val_accuracy: 0.2344 Epoch 78/100 25/25 [==============================] - 37s 1s/step - loss: 0.5707 - accuracy: 0.7625 - val_loss: 2.4277 - val_accuracy: 0.2240 Epoch 79/100 25/25 [==============================] - 36s 1s/step - loss: 0.6518 - accuracy: 0.7300 - val_loss: 1.9735 - val_accuracy: 0.2448 Epoch 80/100 25/25 [==============================] - 36s 1s/step - loss: 0.5680 - accuracy: 0.7613 - val_loss: 1.8923 - val_accuracy: 0.2760 Epoch 81/100 25/25 [==============================] - 36s 1s/step - loss: 0.6130 - accuracy: 0.7412 - val_loss: 1.9575 - val_accuracy: 0.2812 Epoch 82/100 25/25 [==============================] - 36s 1s/step - loss: 0.5543 - accuracy: 0.7750 - val_loss: 1.9804 - val_accuracy: 0.2656 Epoch 83/100 25/25 [==============================] - 36s 1s/step - loss: 0.5484 - accuracy: 0.7588 - val_loss: 2.0896 - val_accuracy: 0.2552 Epoch 84/100 25/25 [==============================] - 36s 1s/step - loss: 0.5875 - accuracy: 0.7500 - val_loss: 2.1894 - val_accuracy: 0.2344 Epoch 85/100 25/25 [==============================] - 36s 1s/step - loss: 0.5803 - accuracy: 0.7588 - val_loss: 2.0186 - val_accuracy: 0.2240 Epoch 86/100 25/25 [==============================] - 36s 1s/step - loss: 0.5683 - accuracy: 0.7688 - val_loss: 2.3338 - val_accuracy: 0.2344 Epoch 87/100 25/25 [==============================] - 36s 1s/step - loss: 0.5192 - accuracy: 0.7613 - val_loss: 2.2686 - val_accuracy: 0.2240 Epoch 88/100 25/25 [==============================] - 36s 1s/step - loss: 0.4986 - accuracy: 0.7713 - val_loss: 2.2567 - val_accuracy: 0.2344 Epoch 89/100 25/25 [==============================] - 36s 1s/step - loss: 0.4916 - accuracy: 0.7925 - val_loss: 2.4030 - val_accuracy: 0.2344 Epoch 90/100 25/25 [==============================] - 36s 1s/step - loss: 0.5215 - accuracy: 0.7700 - val_loss: 2.1048 - val_accuracy: 0.2396 Epoch 91/100 25/25 [==============================] - 36s 1s/step - loss: 0.4871 - accuracy: 0.7850 - val_loss: 2.4577 - val_accuracy: 0.2292 Epoch 92/100 25/25 [==============================] - 36s 1s/step - loss: 0.5142 - accuracy: 0.7688 - val_loss: 2.3448 - val_accuracy: 0.2344 Epoch 93/100 25/25 [==============================] - 36s 1s/step - loss: 0.5245 - accuracy: 0.7837 - val_loss: 2.6430 - val_accuracy: 0.2083 Epoch 94/100 25/25 [==============================] - 37s 1s/step - loss: 0.4893 - accuracy: 0.7700 - val_loss: 2.2934 - val_accuracy: 0.2865 Epoch 95/100 25/25 [==============================] - 39s 2s/step - loss: 0.5475 - accuracy: 0.7675 - val_loss: 1.9955 - val_accuracy: 0.2656 Epoch 96/100 25/25 [==============================] - 37s 1s/step - loss: 0.4791 - accuracy: 0.7900 - val_loss: 2.2937 - val_accuracy: 0.2448 Epoch 97/100 25/25 [==============================] - 37s 1s/step - loss: 0.5287 - accuracy: 0.7750 - val_loss: 2.1882 - val_accuracy: 0.2344 Epoch 98/100 25/25 [==============================] - 36s 1s/step - loss: 0.4331 - accuracy: 0.8150 - val_loss: 2.3296 - val_accuracy: 0.2396 Epoch 99/100 25/25 [==============================] - 37s 1s/step - loss: 0.4675 - accuracy: 0.8000 - val_loss: 2.7121 - val_accuracy: 0.2240 Epoch 100/100 25/25 [==============================] - 36s 1s/step - loss: 0.5437 - accuracy: 0.7738 - val_loss: 2.6578 - val_accuracy: 0.2292
<keras.callbacks.History at 0x2b049b91670>
model_drop.evaluate(test_ds)8/8 [==============================] - 3s 269ms/step - loss: 2.6610 - accuracy: 0.2227
[2.6609723567962646, 0.22265625]
Batch Regularization
Bez dropoutu
model_batch = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.BatchNormalization(),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dense(10, activation='softmax')
])model_batch.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_batch.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_7"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_35 (Conv2D) (None, 55, 55, 96) 34944
batch_normalization (BatchN (None, 55, 55, 96) 384
ormalization)
max_pooling2d_21 (MaxPoolin (None, 27, 27, 96) 0
g2D)
conv2d_36 (Conv2D) (None, 27, 27, 256) 614656
batch_normalization_1 (Batc (None, 27, 27, 256) 1024
hNormalization)
max_pooling2d_22 (MaxPoolin (None, 13, 13, 256) 0
g2D)
conv2d_37 (Conv2D) (None, 13, 13, 384) 885120
batch_normalization_2 (Batc (None, 13, 13, 384) 1536
hNormalization)
conv2d_38 (Conv2D) (None, 13, 13, 384) 1327488
batch_normalization_3 (Batc (None, 13, 13, 384) 1536
hNormalization)
conv2d_39 (Conv2D) (None, 13, 13, 256) 884992
batch_normalization_4 (Batc (None, 13, 13, 256) 1024
hNormalization)
max_pooling2d_23 (MaxPoolin (None, 6, 6, 256) 0
g2D)
flatten_7 (Flatten) (None, 9216) 0
dense_21 (Dense) (None, 4096) 37752832
dense_22 (Dense) (None, 4096) 16781312
dense_23 (Dense) (None, 10) 40970
=================================================================
Total params: 58,327,818
Trainable params: 58,325,066
Non-trainable params: 2,752
_________________________________________________________________
model_batch.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 40s 1s/step - loss: 3.1972 - accuracy: 0.5163 - val_loss: 2.0980 - val_accuracy: 0.1979 Epoch 2/100 25/25 [==============================] - 37s 1s/step - loss: 0.4927 - accuracy: 0.8238 - val_loss: 2.2667 - val_accuracy: 0.1823 Epoch 3/100 25/25 [==============================] - 37s 1s/step - loss: 0.2552 - accuracy: 0.9150 - val_loss: 2.7730 - val_accuracy: 0.1771 Epoch 4/100 25/25 [==============================] - 36s 1s/step - loss: 0.1681 - accuracy: 0.9475 - val_loss: 3.5623 - val_accuracy: 0.1719 Epoch 5/100 25/25 [==============================] - 36s 1s/step - loss: 0.0812 - accuracy: 0.9837 - val_loss: 4.0812 - val_accuracy: 0.1667 Epoch 6/100 25/25 [==============================] - 36s 1s/step - loss: 0.0845 - accuracy: 0.9825 - val_loss: 4.5273 - val_accuracy: 0.1771 Epoch 7/100 25/25 [==============================] - 36s 1s/step - loss: 0.0401 - accuracy: 0.9962 - val_loss: 5.9585 - val_accuracy: 0.1823 Epoch 8/100 25/25 [==============================] - 36s 1s/step - loss: 0.0248 - accuracy: 0.9987 - val_loss: 6.4777 - val_accuracy: 0.1875 Epoch 9/100 25/25 [==============================] - 37s 1s/step - loss: 0.0303 - accuracy: 0.9950 - val_loss: 6.8994 - val_accuracy: 0.1979 Epoch 10/100 25/25 [==============================] - 37s 1s/step - loss: 0.0212 - accuracy: 0.9987 - val_loss: 6.7732 - val_accuracy: 0.2708 Epoch 11/100 25/25 [==============================] - 39s 2s/step - loss: 0.0169 - accuracy: 0.9975 - val_loss: 8.2072 - val_accuracy: 0.2135 Epoch 12/100 25/25 [==============================] - 39s 2s/step - loss: 0.0134 - accuracy: 1.0000 - val_loss: 8.3724 - val_accuracy: 0.2396 Epoch 13/100 25/25 [==============================] - 36s 1s/step - loss: 0.0118 - accuracy: 0.9987 - val_loss: 8.3638 - val_accuracy: 0.2812 Epoch 14/100 25/25 [==============================] - 36s 1s/step - loss: 0.0092 - accuracy: 1.0000 - val_loss: 8.3157 - val_accuracy: 0.2812 Epoch 15/100 25/25 [==============================] - 36s 1s/step - loss: 0.0068 - accuracy: 1.0000 - val_loss: 8.0873 - val_accuracy: 0.3021 Epoch 16/100 25/25 [==============================] - 37s 1s/step - loss: 0.0070 - accuracy: 1.0000 - val_loss: 7.9629 - val_accuracy: 0.2969 Epoch 17/100 25/25 [==============================] - 36s 1s/step - loss: 0.0056 - accuracy: 1.0000 - val_loss: 6.8316 - val_accuracy: 0.3281 Epoch 18/100 25/25 [==============================] - 37s 1s/step - loss: 0.0048 - accuracy: 1.0000 - val_loss: 6.1176 - val_accuracy: 0.3385 Epoch 19/100 25/25 [==============================] - 37s 1s/step - loss: 0.0095 - accuracy: 0.9987 - val_loss: 4.9330 - val_accuracy: 0.3333 Epoch 20/100 25/25 [==============================] - 36s 1s/step - loss: 0.0088 - accuracy: 1.0000 - val_loss: 4.0413 - val_accuracy: 0.4271 Epoch 21/100 25/25 [==============================] - 36s 1s/step - loss: 0.0061 - accuracy: 1.0000 - val_loss: 3.4398 - val_accuracy: 0.4427 Epoch 22/100 25/25 [==============================] - 35s 1s/step - loss: 0.0046 - accuracy: 1.0000 - val_loss: 2.4394 - val_accuracy: 0.5208 Epoch 23/100 25/25 [==============================] - 35s 1s/step - loss: 0.0033 - accuracy: 1.0000 - val_loss: 1.8135 - val_accuracy: 0.5990 Epoch 24/100 25/25 [==============================] - 36s 1s/step - loss: 0.0042 - accuracy: 1.0000 - val_loss: 1.4458 - val_accuracy: 0.6823 Epoch 25/100 25/25 [==============================] - 36s 1s/step - loss: 0.0036 - accuracy: 1.0000 - val_loss: 1.0700 - val_accuracy: 0.7500 Epoch 26/100 25/25 [==============================] - 36s 1s/step - loss: 0.0031 - accuracy: 1.0000 - val_loss: 0.6903 - val_accuracy: 0.8385 Epoch 27/100 25/25 [==============================] - 36s 1s/step - loss: 0.0047 - accuracy: 1.0000 - val_loss: 0.2748 - val_accuracy: 0.9010 Epoch 28/100 25/25 [==============================] - 36s 1s/step - loss: 0.0060 - accuracy: 1.0000 - val_loss: 0.6901 - val_accuracy: 0.8229 Epoch 29/100 25/25 [==============================] - 36s 1s/step - loss: 0.0050 - accuracy: 1.0000 - val_loss: 0.3001 - val_accuracy: 0.9115 Epoch 30/100 25/25 [==============================] - 36s 1s/step - loss: 0.0039 - accuracy: 1.0000 - val_loss: 0.2927 - val_accuracy: 0.9167 Epoch 31/100 25/25 [==============================] - 36s 1s/step - loss: 0.0037 - accuracy: 1.0000 - val_loss: 0.2734 - val_accuracy: 0.9323 Epoch 32/100 25/25 [==============================] - 36s 1s/step - loss: 0.0028 - accuracy: 1.0000 - val_loss: 0.2591 - val_accuracy: 0.9271 Epoch 33/100 25/25 [==============================] - 36s 1s/step - loss: 0.0063 - accuracy: 0.9987 - val_loss: 0.2963 - val_accuracy: 0.9115 Epoch 34/100 25/25 [==============================] - 36s 1s/step - loss: 0.0033 - accuracy: 1.0000 - val_loss: 0.2725 - val_accuracy: 0.9219 Epoch 35/100 25/25 [==============================] - 36s 1s/step - loss: 0.0024 - accuracy: 1.0000 - val_loss: 0.2243 - val_accuracy: 0.9427 Epoch 36/100 25/25 [==============================] - 36s 1s/step - loss: 0.0031 - accuracy: 1.0000 - val_loss: 0.2625 - val_accuracy: 0.9375 Epoch 37/100 25/25 [==============================] - 36s 1s/step - loss: 0.0022 - accuracy: 1.0000 - val_loss: 0.2448 - val_accuracy: 0.9271 Epoch 38/100 25/25 [==============================] - 36s 1s/step - loss: 0.0028 - accuracy: 1.0000 - val_loss: 0.2700 - val_accuracy: 0.9010 Epoch 39/100 25/25 [==============================] - 36s 1s/step - loss: 0.0053 - accuracy: 1.0000 - val_loss: 0.2650 - val_accuracy: 0.9167 Epoch 40/100 25/25 [==============================] - 36s 1s/step - loss: 0.0030 - accuracy: 1.0000 - val_loss: 0.2695 - val_accuracy: 0.9167 Epoch 41/100 25/25 [==============================] - 36s 1s/step - loss: 0.0019 - accuracy: 1.0000 - val_loss: 0.2012 - val_accuracy: 0.9375 Epoch 42/100 25/25 [==============================] - 36s 1s/step - loss: 0.0020 - accuracy: 1.0000 - val_loss: 0.2457 - val_accuracy: 0.9271 Epoch 43/100 25/25 [==============================] - 36s 1s/step - loss: 0.0017 - accuracy: 1.0000 - val_loss: 0.2456 - val_accuracy: 0.9271 Epoch 44/100 25/25 [==============================] - 36s 1s/step - loss: 0.0020 - accuracy: 1.0000 - val_loss: 0.2094 - val_accuracy: 0.9323 Epoch 45/100 25/25 [==============================] - 36s 1s/step - loss: 0.0013 - accuracy: 1.0000 - val_loss: 0.2487 - val_accuracy: 0.9167 Epoch 46/100 25/25 [==============================] - 36s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2507 - val_accuracy: 0.9167 Epoch 47/100 25/25 [==============================] - 36s 1s/step - loss: 0.0015 - accuracy: 1.0000 - val_loss: 0.2457 - val_accuracy: 0.9167 Epoch 48/100 25/25 [==============================] - 37s 1s/step - loss: 0.0019 - accuracy: 1.0000 - val_loss: 0.1729 - val_accuracy: 0.9375 Epoch 49/100 25/25 [==============================] - 36s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2499 - val_accuracy: 0.9167 Epoch 50/100 25/25 [==============================] - 36s 1s/step - loss: 0.0013 - accuracy: 1.0000 - val_loss: 0.2496 - val_accuracy: 0.9271 Epoch 51/100 25/25 [==============================] - 36s 1s/step - loss: 9.5720e-04 - accuracy: 1.0000 - val_loss: 0.2233 - val_accuracy: 0.9375 Epoch 52/100 25/25 [==============================] - 36s 1s/step - loss: 0.0015 - accuracy: 1.0000 - val_loss: 0.2813 - val_accuracy: 0.9219 Epoch 53/100 25/25 [==============================] - 36s 1s/step - loss: 9.6567e-04 - accuracy: 1.0000 - val_loss: 0.2644 - val_accuracy: 0.9219 Epoch 54/100 25/25 [==============================] - 36s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2469 - val_accuracy: 0.9271 Epoch 55/100 25/25 [==============================] - 36s 1s/step - loss: 0.0010 - accuracy: 1.0000 - val_loss: 0.2527 - val_accuracy: 0.9219 Epoch 56/100 25/25 [==============================] - 36s 1s/step - loss: 8.3443e-04 - accuracy: 1.0000 - val_loss: 0.2546 - val_accuracy: 0.9167 Epoch 57/100 25/25 [==============================] - 36s 1s/step - loss: 0.0145 - accuracy: 0.9950 - val_loss: 0.2987 - val_accuracy: 0.9115 Epoch 58/100 25/25 [==============================] - 36s 1s/step - loss: 0.0025 - accuracy: 1.0000 - val_loss: 0.2696 - val_accuracy: 0.9115 Epoch 59/100 25/25 [==============================] - 37s 1s/step - loss: 0.0016 - accuracy: 1.0000 - val_loss: 0.2572 - val_accuracy: 0.9271 Epoch 60/100 25/25 [==============================] - 37s 1s/step - loss: 0.0014 - accuracy: 1.0000 - val_loss: 0.2536 - val_accuracy: 0.9167 Epoch 61/100 25/25 [==============================] - 36s 1s/step - loss: 0.0024 - accuracy: 1.0000 - val_loss: 0.2361 - val_accuracy: 0.9271 Epoch 62/100 25/25 [==============================] - 36s 1s/step - loss: 0.0016 - accuracy: 1.0000 - val_loss: 0.2345 - val_accuracy: 0.9271 Epoch 63/100 25/25 [==============================] - 36s 1s/step - loss: 0.0011 - accuracy: 1.0000 - val_loss: 0.2381 - val_accuracy: 0.9219 Epoch 64/100 25/25 [==============================] - 36s 1s/step - loss: 0.0015 - accuracy: 1.0000 - val_loss: 0.1838 - val_accuracy: 0.9427 Epoch 65/100 25/25 [==============================] - 38s 2s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2077 - val_accuracy: 0.9271 Epoch 66/100 25/25 [==============================] - 36s 1s/step - loss: 9.0273e-04 - accuracy: 1.0000 - val_loss: 0.2517 - val_accuracy: 0.9219 Epoch 67/100 25/25 [==============================] - 36s 1s/step - loss: 9.5187e-04 - accuracy: 1.0000 - val_loss: 0.2012 - val_accuracy: 0.9219 Epoch 68/100 25/25 [==============================] - 36s 1s/step - loss: 9.6461e-04 - accuracy: 1.0000 - val_loss: 0.2327 - val_accuracy: 0.9323 Epoch 69/100 25/25 [==============================] - 37s 1s/step - loss: 9.4358e-04 - accuracy: 1.0000 - val_loss: 0.2500 - val_accuracy: 0.9271 Epoch 70/100 25/25 [==============================] - 37s 1s/step - loss: 0.0013 - accuracy: 1.0000 - val_loss: 0.3345 - val_accuracy: 0.8698 Epoch 71/100 25/25 [==============================] - 36s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2188 - val_accuracy: 0.9271 Epoch 72/100 25/25 [==============================] - 36s 1s/step - loss: 6.7185e-04 - accuracy: 1.0000 - val_loss: 0.2477 - val_accuracy: 0.9271 Epoch 73/100 25/25 [==============================] - 36s 1s/step - loss: 0.0019 - accuracy: 1.0000 - val_loss: 0.2466 - val_accuracy: 0.9167 Epoch 74/100 25/25 [==============================] - 36s 1s/step - loss: 8.7229e-04 - accuracy: 1.0000 - val_loss: 0.2320 - val_accuracy: 0.9115 Epoch 75/100 25/25 [==============================] - 36s 1s/step - loss: 6.4097e-04 - accuracy: 1.0000 - val_loss: 0.2350 - val_accuracy: 0.9115 Epoch 76/100 25/25 [==============================] - 36s 1s/step - loss: 6.8064e-04 - accuracy: 1.0000 - val_loss: 0.2128 - val_accuracy: 0.9323 Epoch 77/100 25/25 [==============================] - 36s 1s/step - loss: 0.0013 - accuracy: 1.0000 - val_loss: 0.2293 - val_accuracy: 0.9323 Epoch 78/100 25/25 [==============================] - 36s 1s/step - loss: 6.8025e-04 - accuracy: 1.0000 - val_loss: 0.2027 - val_accuracy: 0.9271 Epoch 79/100 25/25 [==============================] - 36s 1s/step - loss: 7.1451e-04 - accuracy: 1.0000 - val_loss: 0.2372 - val_accuracy: 0.9219 Epoch 80/100 25/25 [==============================] - 36s 1s/step - loss: 8.0297e-04 - accuracy: 1.0000 - val_loss: 0.2419 - val_accuracy: 0.9219 Epoch 81/100 25/25 [==============================] - 36s 1s/step - loss: 5.7819e-04 - accuracy: 1.0000 - val_loss: 0.2482 - val_accuracy: 0.9115 Epoch 82/100 25/25 [==============================] - 36s 1s/step - loss: 5.6841e-04 - accuracy: 1.0000 - val_loss: 0.2458 - val_accuracy: 0.9219 Epoch 83/100 25/25 [==============================] - 36s 1s/step - loss: 6.9683e-04 - accuracy: 1.0000 - val_loss: 0.2376 - val_accuracy: 0.9219 Epoch 84/100 25/25 [==============================] - 37s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2349 - val_accuracy: 0.9167 Epoch 85/100 25/25 [==============================] - 36s 1s/step - loss: 8.7577e-04 - accuracy: 1.0000 - val_loss: 0.2515 - val_accuracy: 0.9219 Epoch 86/100 25/25 [==============================] - 36s 1s/step - loss: 0.0011 - accuracy: 1.0000 - val_loss: 0.2352 - val_accuracy: 0.9271 Epoch 87/100 25/25 [==============================] - 36s 1s/step - loss: 8.1894e-04 - accuracy: 1.0000 - val_loss: 0.2092 - val_accuracy: 0.9271 Epoch 88/100 25/25 [==============================] - 36s 1s/step - loss: 6.5846e-04 - accuracy: 1.0000 - val_loss: 0.2377 - val_accuracy: 0.9271 Epoch 89/100 25/25 [==============================] - 36s 1s/step - loss: 4.9351e-04 - accuracy: 1.0000 - val_loss: 0.2482 - val_accuracy: 0.9219 Epoch 90/100 25/25 [==============================] - 36s 1s/step - loss: 5.2903e-04 - accuracy: 1.0000 - val_loss: 0.2308 - val_accuracy: 0.9323 Epoch 91/100 25/25 [==============================] - 36s 1s/step - loss: 6.2774e-04 - accuracy: 1.0000 - val_loss: 0.2199 - val_accuracy: 0.9323 Epoch 92/100 25/25 [==============================] - 36s 1s/step - loss: 0.0012 - accuracy: 1.0000 - val_loss: 0.2666 - val_accuracy: 0.9219 Epoch 93/100 25/25 [==============================] - 36s 1s/step - loss: 7.7254e-04 - accuracy: 1.0000 - val_loss: 0.2651 - val_accuracy: 0.9219 Epoch 94/100 25/25 [==============================] - 36s 1s/step - loss: 7.5749e-04 - accuracy: 1.0000 - val_loss: 0.2622 - val_accuracy: 0.9219 Epoch 95/100 25/25 [==============================] - 36s 1s/step - loss: 5.5629e-04 - accuracy: 1.0000 - val_loss: 0.2442 - val_accuracy: 0.9219 Epoch 96/100 25/25 [==============================] - 36s 1s/step - loss: 6.3401e-04 - accuracy: 1.0000 - val_loss: 0.2543 - val_accuracy: 0.9167 Epoch 97/100 25/25 [==============================] - 36s 1s/step - loss: 6.4368e-04 - accuracy: 1.0000 - val_loss: 0.2596 - val_accuracy: 0.9271 Epoch 98/100 25/25 [==============================] - 36s 1s/step - loss: 4.0174e-04 - accuracy: 1.0000 - val_loss: 0.2583 - val_accuracy: 0.9271 Epoch 99/100 25/25 [==============================] - 36s 1s/step - loss: 7.2224e-04 - accuracy: 1.0000 - val_loss: 0.2414 - val_accuracy: 0.9323 Epoch 100/100 25/25 [==============================] - 36s 1s/step - loss: 4.4288e-04 - accuracy: 1.0000 - val_loss: 0.2390 - val_accuracy: 0.9271
<keras.callbacks.History at 0x2b04e37a190>
model_batch.evaluate(test_ds)8/8 [==============================] - 5s 294ms/step - loss: 0.2599 - accuracy: 0.9180
[0.25985682010650635, 0.91796875]
Z dropoutem
model_batch_drop = keras.models.Sequential([
keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=(4,4), activation='relu', input_shape=(227,227,3)),
keras.layers.BatchNormalization(),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Dropout(.5),
keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=(1,1), activation='relu', padding="same"),
keras.layers.BatchNormalization(),
keras.layers.Dropout(.5),
keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2)),
keras.layers.Dropout(.5),
keras.layers.Flatten(),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(4096, activation='relu'),
keras.layers.Dropout(.5),
keras.layers.Dense(10, activation='softmax')
])model_batch_drop.compile(loss='sparse_categorical_crossentropy', optimizer=tf.optimizers.SGD(lr=.001), metrics=['accuracy'])
model_batch_drop.summary()WARNING:absl:`lr` is deprecated, please use `learning_rate` instead, or use the legacy optimizer, e.g.,tf.keras.optimizers.legacy.SGD.
Model: "sequential_8"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_40 (Conv2D) (None, 55, 55, 96) 34944
batch_normalization_5 (Batc (None, 55, 55, 96) 384
hNormalization)
dropout_40 (Dropout) (None, 55, 55, 96) 0
max_pooling2d_24 (MaxPoolin (None, 27, 27, 96) 0
g2D)
dropout_41 (Dropout) (None, 27, 27, 96) 0
conv2d_41 (Conv2D) (None, 27, 27, 256) 614656
batch_normalization_6 (Batc (None, 27, 27, 256) 1024
hNormalization)
dropout_42 (Dropout) (None, 27, 27, 256) 0
max_pooling2d_25 (MaxPoolin (None, 13, 13, 256) 0
g2D)
dropout_43 (Dropout) (None, 13, 13, 256) 0
conv2d_42 (Conv2D) (None, 13, 13, 384) 885120
batch_normalization_7 (Batc (None, 13, 13, 384) 1536
hNormalization)
dropout_44 (Dropout) (None, 13, 13, 384) 0
conv2d_43 (Conv2D) (None, 13, 13, 384) 1327488
batch_normalization_8 (Batc (None, 13, 13, 384) 1536
hNormalization)
dropout_45 (Dropout) (None, 13, 13, 384) 0
conv2d_44 (Conv2D) (None, 13, 13, 256) 884992
batch_normalization_9 (Batc (None, 13, 13, 256) 1024
hNormalization)
dropout_46 (Dropout) (None, 13, 13, 256) 0
max_pooling2d_26 (MaxPoolin (None, 6, 6, 256) 0
g2D)
dropout_47 (Dropout) (None, 6, 6, 256) 0
flatten_8 (Flatten) (None, 9216) 0
dense_24 (Dense) (None, 4096) 37752832
dropout_48 (Dropout) (None, 4096) 0
dense_25 (Dense) (None, 4096) 16781312
dropout_49 (Dropout) (None, 4096) 0
dense_26 (Dense) (None, 10) 40970
=================================================================
Total params: 58,327,818
Trainable params: 58,325,066
Non-trainable params: 2,752
_________________________________________________________________
model_batch_drop.fit(train_ds,
epochs=100,
validation_data=validation_ds,
validation_freq=1,
callbacks=[tensorboard_cb])Epoch 1/100 25/25 [==============================] - 42s 2s/step - loss: 18.7754 - accuracy: 0.2300 - val_loss: 3.0447 - val_accuracy: 0.2500 Epoch 2/100 25/25 [==============================] - 40s 2s/step - loss: 5.7450 - accuracy: 0.2862 - val_loss: 2.0106 - val_accuracy: 0.2031 Epoch 3/100 25/25 [==============================] - 40s 2s/step - loss: 5.0484 - accuracy: 0.2800 - val_loss: 1.8900 - val_accuracy: 0.1927 Epoch 4/100 25/25 [==============================] - 40s 2s/step - loss: 3.9955 - accuracy: 0.2988 - val_loss: 1.8577 - val_accuracy: 0.1979 Epoch 5/100 25/25 [==============================] - 40s 2s/step - loss: 4.0152 - accuracy: 0.3063 - val_loss: 1.8857 - val_accuracy: 0.1979 Epoch 6/100 25/25 [==============================] - 40s 2s/step - loss: 3.2712 - accuracy: 0.3063 - val_loss: 1.8446 - val_accuracy: 0.1250 Epoch 7/100 25/25 [==============================] - 41s 2s/step - loss: 2.5423 - accuracy: 0.3587 - val_loss: 1.8951 - val_accuracy: 0.1875 Epoch 8/100 25/25 [==============================] - 44s 2s/step - loss: 2.3186 - accuracy: 0.3625 - val_loss: 1.8989 - val_accuracy: 0.1615 Epoch 9/100 25/25 [==============================] - 42s 2s/step - loss: 2.1973 - accuracy: 0.3663 - val_loss: 2.0297 - val_accuracy: 0.2031 Epoch 10/100 25/25 [==============================] - 40s 2s/step - loss: 1.8098 - accuracy: 0.4275 - val_loss: 2.0467 - val_accuracy: 0.2083 Epoch 11/100 25/25 [==============================] - 40s 2s/step - loss: 1.6218 - accuracy: 0.4888 - val_loss: 2.1542 - val_accuracy: 0.1615 Epoch 12/100 25/25 [==============================] - 40s 2s/step - loss: 1.7138 - accuracy: 0.4762 - val_loss: 2.4840 - val_accuracy: 0.2292 Epoch 13/100 25/25 [==============================] - 40s 2s/step - loss: 1.5391 - accuracy: 0.5238 - val_loss: 2.5269 - val_accuracy: 0.1667 Epoch 14/100 25/25 [==============================] - 41s 2s/step - loss: 1.3664 - accuracy: 0.5763 - val_loss: 2.6138 - val_accuracy: 0.2188 Epoch 15/100 25/25 [==============================] - 40s 2s/step - loss: 1.3573 - accuracy: 0.5537 - val_loss: 3.0465 - val_accuracy: 0.2083 Epoch 16/100 25/25 [==============================] - 40s 2s/step - loss: 1.3332 - accuracy: 0.5638 - val_loss: 2.8962 - val_accuracy: 0.2240 Epoch 17/100 25/25 [==============================] - 40s 2s/step - loss: 1.2500 - accuracy: 0.5863 - val_loss: 2.7711 - val_accuracy: 0.2188 Epoch 18/100 25/25 [==============================] - 40s 2s/step - loss: 1.1243 - accuracy: 0.6100 - val_loss: 2.8280 - val_accuracy: 0.2240 Epoch 19/100 25/25 [==============================] - 40s 2s/step - loss: 1.0187 - accuracy: 0.6150 - val_loss: 2.7781 - val_accuracy: 0.2240 Epoch 20/100 25/25 [==============================] - 40s 2s/step - loss: 1.0591 - accuracy: 0.6263 - val_loss: 2.7976 - val_accuracy: 0.2240 Epoch 21/100 25/25 [==============================] - 40s 2s/step - loss: 0.9730 - accuracy: 0.6513 - val_loss: 2.6593 - val_accuracy: 0.2500 Epoch 22/100 25/25 [==============================] - 40s 2s/step - loss: 1.0428 - accuracy: 0.6538 - val_loss: 2.9169 - val_accuracy: 0.2396 Epoch 23/100 25/25 [==============================] - 41s 2s/step - loss: 0.9658 - accuracy: 0.6587 - val_loss: 2.9805 - val_accuracy: 0.2396 Epoch 24/100 25/25 [==============================] - 40s 2s/step - loss: 0.8756 - accuracy: 0.6625 - val_loss: 2.9590 - val_accuracy: 0.2604 Epoch 25/100 25/25 [==============================] - 40s 2s/step - loss: 0.9395 - accuracy: 0.6737 - val_loss: 2.7804 - val_accuracy: 0.2188 Epoch 26/100 25/25 [==============================] - 40s 2s/step - loss: 0.8502 - accuracy: 0.6938 - val_loss: 2.9480 - val_accuracy: 0.2448 Epoch 27/100 25/25 [==============================] - 40s 2s/step - loss: 0.8092 - accuracy: 0.7350 - val_loss: 3.3202 - val_accuracy: 0.2188 Epoch 28/100 25/25 [==============================] - 40s 2s/step - loss: 0.7796 - accuracy: 0.7150 - val_loss: 2.8618 - val_accuracy: 0.1823 Epoch 29/100 25/25 [==============================] - 40s 2s/step - loss: 0.7650 - accuracy: 0.7175 - val_loss: 3.1987 - val_accuracy: 0.1719 Epoch 30/100 25/25 [==============================] - 40s 2s/step - loss: 0.6968 - accuracy: 0.7437 - val_loss: 3.0807 - val_accuracy: 0.1562 Epoch 31/100 25/25 [==============================] - 40s 2s/step - loss: 0.7408 - accuracy: 0.7387 - val_loss: 4.7095 - val_accuracy: 0.1771 Epoch 32/100 25/25 [==============================] - 40s 2s/step - loss: 0.7883 - accuracy: 0.7138 - val_loss: 3.5988 - val_accuracy: 0.1250 Epoch 33/100 25/25 [==============================] - 40s 2s/step - loss: 0.6965 - accuracy: 0.7237 - val_loss: 3.8168 - val_accuracy: 0.1771 Epoch 34/100 25/25 [==============================] - 42s 2s/step - loss: 0.6220 - accuracy: 0.7725 - val_loss: 3.1421 - val_accuracy: 0.1615 Epoch 35/100 25/25 [==============================] - 40s 2s/step - loss: 0.6952 - accuracy: 0.7563 - val_loss: 3.3574 - val_accuracy: 0.1771 Epoch 36/100 25/25 [==============================] - 40s 2s/step - loss: 0.6311 - accuracy: 0.7638 - val_loss: 3.9404 - val_accuracy: 0.1250 Epoch 37/100 25/25 [==============================] - 40s 2s/step - loss: 0.6131 - accuracy: 0.7713 - val_loss: 4.1547 - val_accuracy: 0.1094 Epoch 38/100 25/25 [==============================] - 40s 2s/step - loss: 0.6917 - accuracy: 0.7575 - val_loss: 3.6349 - val_accuracy: 0.1562 Epoch 39/100 25/25 [==============================] - 40s 2s/step - loss: 0.6401 - accuracy: 0.7825 - val_loss: 3.6254 - val_accuracy: 0.1458 Epoch 40/100 25/25 [==============================] - 40s 2s/step - loss: 0.6787 - accuracy: 0.7437 - val_loss: 3.9394 - val_accuracy: 0.1562 Epoch 41/100 25/25 [==============================] - 40s 2s/step - loss: 0.5985 - accuracy: 0.7825 - val_loss: 3.8814 - val_accuracy: 0.1562 Epoch 42/100 25/25 [==============================] - 40s 2s/step - loss: 0.5920 - accuracy: 0.7850 - val_loss: 4.7182 - val_accuracy: 0.1302 Epoch 43/100 25/25 [==============================] - 40s 2s/step - loss: 0.6628 - accuracy: 0.7563 - val_loss: 3.4734 - val_accuracy: 0.1719 Epoch 44/100 25/25 [==============================] - 40s 2s/step - loss: 0.5662 - accuracy: 0.7912 - val_loss: 3.6088 - val_accuracy: 0.1667 Epoch 45/100 25/25 [==============================] - 42s 2s/step - loss: 0.5879 - accuracy: 0.7763 - val_loss: 3.9871 - val_accuracy: 0.1302 Epoch 46/100 25/25 [==============================] - 40s 2s/step - loss: 0.4717 - accuracy: 0.8062 - val_loss: 3.6940 - val_accuracy: 0.1719 Epoch 47/100 25/25 [==============================] - 41s 2s/step - loss: 0.5342 - accuracy: 0.8037 - val_loss: 4.1004 - val_accuracy: 0.1615 Epoch 48/100 25/25 [==============================] - 41s 2s/step - loss: 0.5903 - accuracy: 0.7800 - val_loss: 4.2024 - val_accuracy: 0.1667 Epoch 49/100 25/25 [==============================] - 41s 2s/step - loss: 0.6181 - accuracy: 0.7887 - val_loss: 4.1097 - val_accuracy: 0.1458 Epoch 50/100 25/25 [==============================] - 41s 2s/step - loss: 0.5110 - accuracy: 0.8112 - val_loss: 4.3863 - val_accuracy: 0.1979 Epoch 51/100 25/25 [==============================] - 42s 2s/step - loss: 0.5459 - accuracy: 0.8037 - val_loss: 4.2418 - val_accuracy: 0.1354 Epoch 52/100 25/25 [==============================] - 42s 2s/step - loss: 0.5215 - accuracy: 0.8062 - val_loss: 4.4408 - val_accuracy: 0.2135 Epoch 53/100 25/25 [==============================] - 42s 2s/step - loss: 0.5577 - accuracy: 0.8213 - val_loss: 5.2647 - val_accuracy: 0.1510 Epoch 54/100 25/25 [==============================] - 40s 2s/step - loss: 0.5364 - accuracy: 0.7887 - val_loss: 4.3820 - val_accuracy: 0.1615 Epoch 55/100 25/25 [==============================] - 41s 2s/step - loss: 0.5056 - accuracy: 0.8163 - val_loss: 3.0960 - val_accuracy: 0.2292 Epoch 56/100 25/25 [==============================] - 44s 2s/step - loss: 0.5088 - accuracy: 0.8275 - val_loss: 4.1318 - val_accuracy: 0.1667 Epoch 57/100 25/25 [==============================] - 42s 2s/step - loss: 0.5356 - accuracy: 0.7987 - val_loss: 3.8712 - val_accuracy: 0.1562 Epoch 58/100 25/25 [==============================] - 42s 2s/step - loss: 0.4303 - accuracy: 0.8388 - val_loss: 3.7644 - val_accuracy: 0.1771 Epoch 59/100 25/25 [==============================] - 42s 2s/step - loss: 0.3881 - accuracy: 0.8350 - val_loss: 4.0027 - val_accuracy: 0.1719 Epoch 60/100 25/25 [==============================] - 41s 2s/step - loss: 0.5084 - accuracy: 0.8338 - val_loss: 4.0352 - val_accuracy: 0.1458 Epoch 61/100 25/25 [==============================] - 41s 2s/step - loss: 0.4368 - accuracy: 0.8238 - val_loss: 3.6374 - val_accuracy: 0.1615 Epoch 62/100 25/25 [==============================] - 42s 2s/step - loss: 0.4437 - accuracy: 0.8300 - val_loss: 3.5370 - val_accuracy: 0.1979 Epoch 63/100 25/25 [==============================] - 41s 2s/step - loss: 0.4130 - accuracy: 0.8425 - val_loss: 3.9090 - val_accuracy: 0.1458 Epoch 64/100 25/25 [==============================] - 41s 2s/step - loss: 0.3847 - accuracy: 0.8500 - val_loss: 3.9670 - val_accuracy: 0.1510 Epoch 65/100 25/25 [==============================] - 41s 2s/step - loss: 0.4029 - accuracy: 0.8425 - val_loss: 4.5627 - val_accuracy: 0.1302 Epoch 66/100 25/25 [==============================] - 41s 2s/step - loss: 0.3499 - accuracy: 0.8550 - val_loss: 3.7520 - val_accuracy: 0.1667 Epoch 67/100 25/25 [==============================] - 41s 2s/step - loss: 0.3891 - accuracy: 0.8600 - val_loss: 4.5331 - val_accuracy: 0.1771 Epoch 68/100 25/25 [==============================] - 41s 2s/step - loss: 0.4390 - accuracy: 0.8363 - val_loss: 4.4191 - val_accuracy: 0.1823 Epoch 69/100 25/25 [==============================] - 41s 2s/step - loss: 0.3524 - accuracy: 0.8600 - val_loss: 5.4688 - val_accuracy: 0.1406 Epoch 70/100 25/25 [==============================] - 41s 2s/step - loss: 0.4448 - accuracy: 0.8550 - val_loss: 4.3226 - val_accuracy: 0.1927 Epoch 71/100 25/25 [==============================] - 41s 2s/step - loss: 0.3959 - accuracy: 0.8525 - val_loss: 4.6061 - val_accuracy: 0.2031 Epoch 72/100 25/25 [==============================] - 41s 2s/step - loss: 0.3506 - accuracy: 0.8775 - val_loss: 3.8307 - val_accuracy: 0.1615 Epoch 73/100 25/25 [==============================] - 42s 2s/step - loss: 0.3496 - accuracy: 0.8750 - val_loss: 4.3153 - val_accuracy: 0.1823 Epoch 74/100 25/25 [==============================] - 42s 2s/step - loss: 0.4924 - accuracy: 0.8263 - val_loss: 3.8766 - val_accuracy: 0.1875 Epoch 75/100 25/25 [==============================] - 41s 2s/step - loss: 0.4854 - accuracy: 0.8487 - val_loss: 5.7284 - val_accuracy: 0.1562 Epoch 76/100 25/25 [==============================] - 41s 2s/step - loss: 0.4044 - accuracy: 0.8625 - val_loss: 5.3349 - val_accuracy: 0.0938 Epoch 77/100 25/25 [==============================] - 42s 2s/step - loss: 0.4885 - accuracy: 0.8375 - val_loss: 3.9680 - val_accuracy: 0.1719 Epoch 78/100 25/25 [==============================] - 41s 2s/step - loss: 0.3635 - accuracy: 0.8500 - val_loss: 4.1063 - val_accuracy: 0.1719 Epoch 79/100 25/25 [==============================] - 41s 2s/step - loss: 0.3271 - accuracy: 0.8675 - val_loss: 4.6011 - val_accuracy: 0.1615 Epoch 80/100 25/25 [==============================] - 42s 2s/step - loss: 0.3255 - accuracy: 0.8712 - val_loss: 4.8067 - val_accuracy: 0.1771 Epoch 81/100 25/25 [==============================] - 42s 2s/step - loss: 0.3612 - accuracy: 0.8600 - val_loss: 4.4704 - val_accuracy: 0.1562 Epoch 82/100 25/25 [==============================] - 42s 2s/step - loss: 0.3358 - accuracy: 0.8775 - val_loss: 4.3849 - val_accuracy: 0.1562 Epoch 83/100 25/25 [==============================] - 41s 2s/step - loss: 0.3253 - accuracy: 0.8850 - val_loss: 4.4543 - val_accuracy: 0.1927 Epoch 84/100 25/25 [==============================] - 41s 2s/step - loss: 0.2932 - accuracy: 0.8800 - val_loss: 4.2358 - val_accuracy: 0.2188 Epoch 85/100 25/25 [==============================] - 41s 2s/step - loss: 0.2768 - accuracy: 0.8938 - val_loss: 4.6282 - val_accuracy: 0.2500 Epoch 86/100 25/25 [==============================] - 42s 2s/step - loss: 0.3475 - accuracy: 0.8800 - val_loss: 4.8860 - val_accuracy: 0.1979 Epoch 87/100 25/25 [==============================] - 41s 2s/step - loss: 0.4666 - accuracy: 0.8575 - val_loss: 6.6145 - val_accuracy: 0.1719 Epoch 88/100 25/25 [==============================] - 41s 2s/step - loss: 0.3518 - accuracy: 0.8712 - val_loss: 6.0062 - val_accuracy: 0.1615 Epoch 89/100 25/25 [==============================] - 42s 2s/step - loss: 0.3243 - accuracy: 0.8725 - val_loss: 5.5636 - val_accuracy: 0.1979 Epoch 90/100 25/25 [==============================] - 42s 2s/step - loss: 0.3085 - accuracy: 0.8950 - val_loss: 4.9411 - val_accuracy: 0.1562 Epoch 91/100 25/25 [==============================] - 41s 2s/step - loss: 0.3635 - accuracy: 0.8838 - val_loss: 8.3831 - val_accuracy: 0.1510 Epoch 92/100 25/25 [==============================] - 41s 2s/step - loss: 0.3729 - accuracy: 0.8675 - val_loss: 5.3772 - val_accuracy: 0.1458 Epoch 93/100 25/25 [==============================] - 41s 2s/step - loss: 0.4809 - accuracy: 0.8462 - val_loss: 3.2538 - val_accuracy: 0.2031 Epoch 94/100 25/25 [==============================] - 41s 2s/step - loss: 0.3623 - accuracy: 0.8675 - val_loss: 4.4434 - val_accuracy: 0.1875 Epoch 95/100 25/25 [==============================] - 41s 2s/step - loss: 0.3453 - accuracy: 0.8675 - val_loss: 7.2843 - val_accuracy: 0.1771 Epoch 96/100 25/25 [==============================] - 41s 2s/step - loss: 0.4527 - accuracy: 0.8700 - val_loss: 3.5670 - val_accuracy: 0.2240 Epoch 97/100 25/25 [==============================] - 41s 2s/step - loss: 0.3839 - accuracy: 0.8662 - val_loss: 4.9288 - val_accuracy: 0.1458 Epoch 98/100 25/25 [==============================] - 41s 2s/step - loss: 0.2978 - accuracy: 0.8863 - val_loss: 5.4240 - val_accuracy: 0.2292 Epoch 99/100 25/25 [==============================] - 41s 2s/step - loss: 0.3522 - accuracy: 0.8750 - val_loss: 5.5607 - val_accuracy: 0.1771 Epoch 100/100 25/25 [==============================] - 42s 2s/step - loss: 0.3677 - accuracy: 0.8775 - val_loss: 7.4051 - val_accuracy: 0.1146
<keras.callbacks.History at 0x2b05112a220>
model_batch_drop.evaluate(test_ds)8/8 [==============================] - 5s 283ms/step - loss: 9.0089 - accuracy: 0.1758
[9.008861541748047, 0.17578125]