6.Cats.Vs.Dogs.Network.Pretrained.py

from keras.applications import VGG16
from keras.preprocessing.image import ImageDataGenerator
from keras import layers, models, optimizers
import NvidiaMixedPrec, os
import numpy as np

datagen = ImageDataGenerator(rescale=1/255)

working_dir = '/home/andrew/PycharmProjects/DeepLearning/CatDog/CatDogWorking'

train_dir = os.path.join(working_dir, 'train')
val_dir = os.path.join(working_dir, 'val')
test_dir = os.path.join(working_dir, 'test')

conv_base = VGG16(weights='imagenet', include_top=False, input_shape=(150, 150, 3))
batch_size = 20


def extract_features(directory, sample_count):
    features = np.zeros(shape=(sample_count, 4, 4, 512))
    labels = np.zeros(shape=sample_count)
    generator = datagen.flow_from_directory(directory, target_size=(150, 150), batch_size=batch_size, class_mode='binary')
    i = 0
    for inputs_batch, labels_batch in generator:
        features_batch = conv_base.predict(inputs_batch)
        features[i*batch_size:(i+1)*batch_size] = features_batch
        labels[i*batch_size:(i+1)*batch_size] = labels_batch
        i += 1
        if i*batch_size >= sample_count:
            break
    return features, labels


train_features, train_labels = extract_features(train_dir, 2000)
val_features, val_labels = extract_features(val_dir, 1000)
test_features, test_labels = extract_features(test_dir, 1000)

train_features = np.reshape(train_features, (2000, 4*4*512))
val_features = np.reshape(val_features, (1000, 4*4*512))
test_features = np.reshape(test_features, (1000, 4*4*512))

model = models.Sequential()
model.add(layers.Dense(256, activation='relu', input_dim=4*4*512))
model.add(layers.Dropout(.5))
model.add(layers.Dense(1, activation='sigmoid'))

model.compile(optimizer=optimizers.RMSprop(lr=2e-5), loss='binary_crossentropy', metrics=['acc'])

history = model.fit_generator(train_features, train_labels, epochs=30, batch_size=batch_size, validation_data=(val_features, val_labels))