model.load_weights(resumeFrom, by_name=True)
# un-freeze training for a few bottom layers
for layer in model.layers[:39]:
layer.trainable = True
model.compile(
loss=
'mean_absolute_error', # mean_squared_error mean_absolute_error
optimizer=optimizers.Nadam(lr=0.000001),
metrics=[r2_metrics],
decay=0.0005)
# print_summary(model)
history = model.fit_generator(
dataset.generate(batch_size=batch_size,
is_training=True,
max_extra_scale=2.2),
steps_per_epoch=200,
epochs=epochs2,
initial_epoch=epochs1,
validation_data=dataset.generate(batch_size=batch_size,
is_training=False),
validation_steps=len(dataset.test_items),
pickle_safe=True,
verbose=1,
callbacks=[checkpoint, tensorboard])
kir_save_history(history, 'scale_pre')
# metrics=['accuracy'])
# model.compile(loss = "categorical_crossentropy",
# optimizer = optimizers.SGD(lr=0.00003, momentum=0.9, nesterov=True, decay=1e-6),
# metrics=["accuracy"],
# decay=0.0005)
model.compile(loss='categorical_crossentropy',
optimizer=optimizers.Nadam(lr=0.000001),
metrics=['accuracy'])
# model.compile(loss = "mean_absolute_error",
# optimizer = optimizers.SGD(lr=0.0001, momentum=0.9, nesterov=True),
# metrics=["accuracy"],
# decay=0.0005)
print_summary(model)
plot_model(model, to_file='model.pdf', show_shapes=True)
history = model.fit_generator(train_generator,
steps_per_epoch=200,
epochs=epochs1,
validation_data=valid_generator,
validation_steps=50,
initial_epoch=0,
pickle_safe=True,
verbose=1,
callbacks=[checkpoint, tensorboard])
kir_save_history(history, 'pre')
print_summary(model)
plot_model(model, to_file='model.pdf', show_shapes=True)
# ===== first stage
history = model.fit_generator(train_generator,
steps_per_epoch=300,
epochs=epochs1,
validation_data=validation_generator,
validation_steps=50,
initial_epoch=0,
pickle_safe=True,
verbose=1,
callbacks=[checkpoint, tensorboard])
kir_save_history(history, 'crops_1')
else:
print('Resume training after %s\'s epoch' % epochs1)
model = load_model(resumeFrom)
# Unfreeze all layers
for layer in model.layers:
layer.trainable = True
model.compile(loss='categorical_crossentropy',
optimizer=optimizers.Nadam(lr=0.000001),
metrics=['accuracy'],
decay=0.0005)
history = model.fit_generator(train_generator,