VGG16. Классификация изображений (Cats vs. Dogs)

"""
VGG-16 Implementation on Cats&Dogs Dataset
"""
 
import keras
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
from keras.backend.tensorflow_backend import set_session
from keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.layers import Conv2D, Dense, Flatten, MaxPool2D
from tensorflow.keras.models import Sequential, load_model
from keras.optimizers import Adam
from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator
 
config = tf.compat.v1.ConfigProto()
config.gpu_options.allow_growth = True  # dynamically grow the memory used on the GPU
config.log_device_placement = True  # to log device placement (on which device the operation ran)
sess = tf.compat.v1.Session(config=config)
set_session(sess)
 
# Get the data
trdata = ImageDataGenerator()
traindata = trdata.flow_from_directory(directory="C:/Temp/Datasets/CatsDogs/train", target_size=(224, 224))
tsdata = ImageDataGenerator()
testdata = tsdata.flow_from_directory(directory='C:/Temp/Datasets/CatsDogs/validation', target_size=(224, 224))
 
# Generate the model
model = Sequential()
# Layer 1: Convolutional
model.add(Conv2D(input_shape=(224, 224, 3), filters=64, kernel_size=(3, 3),
                 padding='same', activation='relu'))
# Layer 2: Convolutional
model.add(Conv2D(filters=64, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 3: MaxPooling
model.add(MaxPool2D(pool_size=(2, 2), strides=(2, 2)))
 
# Layer 4: Convolutional
model.add(Conv2D(filters=128, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 5: Convolutional
model.add(Conv2D(filters=128, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 6: MaxPooling
model.add(MaxPool2D(pool_size=(2, 2), strides=(2, 2)))
 
# Layer 7: Convolutional
model.add(Conv2D(filters=256, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 8: Convolutional
model.add(Conv2D(filters=256, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 9: Convolutional
model.add(Conv2D(filters=256, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 10: MaxPooling
model.add(MaxPool2D(pool_size=(2, 2), strides=(2, 2)))
 
# Layer 11: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 12: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 13: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 14: MaxPooling
model.add(MaxPool2D(pool_size=(2, 2), strides=(2, 2)))
 
# Layer 15: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 16: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 17: Convolutional
model.add(Conv2D(filters=512, kernel_size=(3, 3), padding='same', activation='relu'))
# Layer 18: MaxPooling
model.add(MaxPool2D(pool_size=(2, 2), strides=(2, 2)))
 
# Layer 19: Flatten
model.add(Flatten())
# Layer 20: Fully Connected Layer
model.add(Dense(units=4096, activation='relu'))
# Layer 21: Fully Connected Layer
model.add(Dense(units=4096, activation='relu'))
# Layer 22: Softmax Layer
model.add(Dense(units=2, activation='softmax'))
 
# Add Optimizer and check accuracy metrics
optimizer = tf.keras.optimizers.Adam(lr=0.001)
model.compile(optimizer=optimizer, loss=keras.losses.categorical_crossentropy,
              metrics=['accuracy'])
# Check model summary
print(model.summary())
 
checkpoint = ModelCheckpoint("vgg16_1.h5", monitor='val_acc', verbose=1, save_best_only=True,
                             save_weights_only=False, mode='auto', period=1)
earlystop = EarlyStopping(monitor='val_acc', min_delta=0, patience=20, verbose=1, mode='auto')
hist = model.fit_generator(steps_per_epoch=100, generator=traindata, validation_data=testdata,
                           validation_steps=10, epochs=100,
                           callbacks=[checkpoint, earlystop])
 
plt.plot(hist.history["acc"])
plt.plot(hist.history['val_acc'])
plt.plot(hist.history['loss'])
plt.plot(hist.history['val_loss'])
plt.title("model accuracy")
plt.ylabel("Accuracy")
plt.xlabel("Epoch")
plt.legend(["Accuracy", "Validation Accuracy", "loss", "Validation Loss"])
plt.show(block=True)
 
# Try on test data
img = image.load_img("C:/Temp/Datasets/CatsDogs/validation/39.jpg", target_size=(224, 224))
img = np.asarray(img)
plt.imshow(img)
img = np.expand_dims(img, axis=0)
saved_model = load_model("vgg16_1.h5")
output = saved_model.predict(img)
if output[0][0] > output[0][1]:
    print("cat")
else:
    print('dog')

optimizer = keras.optimizers.adam_v2.Adam(learning_rate=0.001)
model.compile(optimizer=optimizer, loss=tf.keras.losses.CategoricalCrossentropy,
              metrics=['accuracy'])

print(model.summary())