VGG16FE.py

from keras.preprocessing.image import ImageDataGenerator
from keras.applications.vgg16 import VGG16
from keras import models
from keras import layers
from keras.callbacks import EarlyStopping, ReduceLROnPlateau, ModelCheckpoint, TensorBoard

# locations of training data
train_loc = "C:\\Users\\marcu\\learning\\mats\\train"
test_loc = "C:\\Users\\marcu\\learning\\mats\\test"

# Load CNN from VGG16
conv_base = VGG16(weights='imagenet',include_top = False, input_shape = (224,224,3))
conv_base.trainable = True
set_trainable = False
for layer in conv_base.layers:
    if layer.name == 'block5_conv2':
        set_trainable = True
    if set_trainable:
        layer.trainable = True
    else:
        layer.trainable = False

# Add dense DNN on top
model = models.Sequential()
model.add(conv_base)
model.add(layers.Flatten())
model.add(layers.Dense(256, activation = 'relu'))
model.add(layers.Dense(128, activation = 'relu'))
model.add(layers.Dense(5,activation = 'softmax'))

# Augment training data
train_datagen = ImageDataGenerator(
        rescale = 1./255,
        rotation_range = 180,
        width_shift_range = 0.2,
        height_shift_range = 0.2,
        shear_range = 0,
        zoom_range = .2,
        horizontal_flip = True,
        fill_mode = 'nearest'
        )

# Normalize test data
test_datagen = ImageDataGenerator(rescale=1./255)


train_generator = train_datagen.flow_from_directory(
        train_loc,
        target_size = (224,224),
        batch_size=20,
        class_mode='categorical'
        )

valid_generator = test_datagen.flow_from_directory(
        test_loc,
        target_size = (224,224),
        batch_size = 20,
        class_mode = 'categorical')

model.compile(loss = 'categorical_crossentropy',
              optimizer = 'rmsprop',
              metrics = ['accuracy'])

callbacks = [EarlyStopping(monitor='val_loss',
                           patience=8,
                           verbose=1,
                           min_delta=1e-4),
             ReduceLROnPlateau(monitor='val_loss',
                               factor=0.1,
                               patience=4,
                               verbose=1,
                               epsilon=1e-4),
             ModelCheckpoint(monitor='val_loss',
                             filepath='modelWeights/best_weights_VGG16.hdf5',
                             save_best_only=True,
                             save_weights_only=True),
             TensorBoard(log_dir='logs')]

history = model.fit_generator(
        train_generator,
        steps_per_epoch = 100,
        epochs = 100,
        validation_data = valid_generator,
        callbacks = callbacks,
        validation_steps = 50)