nb_classes = 9
# Load extracted bottleneck features
train_data = np.load(learning_data_root+'inceptionv3_bottleneck_features_train.npy')
validation_data = np.load(learning_data_root+'inceptionv3_bottleneck_features_validation.npy')
train_labels = np.load(learning_data_root+'inceptionv3_bottleneck_labels_training.npy')
validation_labels = np.load(learning_data_root+'inceptionv3_bottleneck_labels_validation.npy')
# Hyperparameters
learning_rates = [0.005, 0.001, 0.0005, 0.0001]
# Hyperparameter search
for lr in learning_rates:
save_string = utility.save_string(0, lr)
utility.log(save_string)
# Create Optimizer
optimizerSGD = tf.keras.optimizers.SGD(lr=lr, momentum=0.9)
optimizerAdam = tf.keras.optimizers.Adam(lr=lr, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0,)
model = TopClassifier().create_model(nb_classes=nb_classes, optimizer=optimizerAdam,input_shape=train_data.shape[1:])
tbCallBack = tf.keras.callbacks.TensorBoard(log_dir='./logs', histogram_freq=0, batch_size=batch_size, write_graph=True,
write_grads=False, write_images=False,
#embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None
)
model.fit(train_data, train_labels,
epochs=epoch_size,
validation_photo_to_label_dict,
batch_size=batch_size,
target_size=(img_width, img_height),
train_or_valid='validation')
validation_generator = validation_multilabel_datagen.flow()
# Hyperparameters
num_freezed_layers_array = [19]
learning_rates = [0.001]
# Hyperparameter search
for num_freezed_layers in num_freezed_layers_array:
for lr in learning_rates:
save_string = utility.save_string(num_freezed_layers, lr)
utility.log(save_string)
# Create Optimizer
optimizerSGD = tf.keras.optimizers.SGD(lr=lr, momentum=0.9)
optimizerAdam = tf.keras.optimizers.Adam(
lr=lr,
beta_1=0.9,
beta_2=0.999,
epsilon=1e-08,
decay=0.0,
)
# Create model
model = VGG16Model().create_model(num_freezedLayers=num_freezed_layers,
nb_classes=nb_classes,