def create_vgg16_model(input_shape):
base_model = vgg16(input_shape, include_top=False)
x = Flatten()(base_model.output)
x = Dense(512, activation="relu")(x)
x = Dropout(0.5)(x)
predictions = Dense(1, activation="sigmoid")(x)
model_name = "vgg16_full_fc512_fc1_aug_reduce_lr"
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape):
base_model = vgg16(input_shape, include_top=False)
block4_layer = base_model.get_layer("block4_pool")
x = Flatten()(block4_layer.output)
x = Dense(512, activation="relu")(x)
x = Dropout(0.5)(x)
predictions = Dense(1, activation="sigmoid")(x)
model_name = "vgg16_block4_pool_fc512_fc1_aug"
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape):
base_model = vgg16(input_shape,
include_top=False,
batch_normalization=False)
x = Flatten()(base_model.output)
x = Dense(512, activation="relu")(x)
predictions = Dense(1, activation="sigmoid")(x)
model_name = "vgg16_full_fc512_fc1_baseline"
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape):
base_model = vgg16(input_shape, weights="imagenet", include_top=False)
for layer in base_model.layers:
layer.trainable = False
x = Flatten()(base_model.output)
x = Dense(512, activation="relu")(x)
x = Dropout(0.5)(x)
predictions = Dense(1, activation="sigmoid")(x)
model_name = "vgg16_pretrained_full_fc512_fc1_aug_reduce_lr"
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape, layer_index=15):
base_model = vgg16(input_shape, include_top=False, batch_normalization=False)
for layer in base_model.layers[:layer_index]:
layer.trainable = False
for layer in base_model.layers[layer_index:]:
layer.trainable = True
x = Flatten()(base_model.output)
l2 = 0.01
predictions = Dense(1, activation="sigmoid", kernel_regularizer=regularizers.l2(l2))(x)
model_name = "vgg16_full_fc1_aug_frozen_rms_prop_l2_{}".format(l2)
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape, layer_index=15):
base_model = vgg16(input_shape,
include_top=False,
batch_normalization=False)
for layer in base_model.layers[:layer_index]:
layer.trainable = False
for layer in base_model.layers[layer_index:]:
layer.trainable = True
x = Flatten()(base_model.output)
predictions = Dense(2, activation="softmax")(x)
model_name = "vgg16_full_fc1_aug_frozen_{}_softmax".format(layer_index)
return model_name, Model(base_model.input, predictions)
def create_vgg16_model(input_shape, layer_index=15):
base_model = vgg16(input_shape, include_top=False, batch_normalization=True)
for layer in base_model.layers[:layer_index]:
layer.trainable = True
for layer in base_model.layers[layer_index:]:
layer.trainable = True
x = Flatten()(base_model.output)
x = Dense(512, activation="relu")(x)
x = Dropout(0.5)(x)
x = Dense(512, activation="relu")(x)
x = Dropout(0.5)(x)
predictions = Dense(1, activation="sigmoid")(x)
model_name = "vgg16_full_fc512_fc512_fc1_aug_frozen_sgd"
return model_name, Model(base_model.input, predictions)