return model
def train_network(model, train_generator, validation_generator):
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-4),
metrics=["accuracy"])
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples / train_generator.batch_size,
epochs=30,
validation_data=validation_generator,
validation_steps=validation_generator.samples /
validation_generator.batch_size,
verbose=2)
return history
if __name__ == '__main__':
preprocessor = DataPreprocessor.DataPreprocessor()
model = built_cnn()
#model = loader.load_model_from_json('3_conv_adam_1e-3_cat_cross_v2.json', '3_conv_adam_1e-3_cat_cross_v2.h5')
train_generator, validation_generator = preprocessor.get_image_generators()
history = train_network(model, train_generator, validation_generator)
loader.save_model(model, '3_conv_adam_1e-3_cat_cross_v2.json',
'3_conv_adam_1e-3_cat_cross_v2.h5')
#plot.plot_model(model, 'cnn_3_conv_RMSProp_1e-4.png')
plot.plot_history(history)
def train_network(model, train_generator, validation_generator):
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=1e-4),
metrics=["accuracy"])
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples / train_generator.batch_size,
epochs=20,
validation_data=validation_generator,
validation_steps=validation_generator.samples /
validation_generator.batch_size,
verbose=2)
return history
if __name__ == '__main__':
preprocessor = DataPreprocessor.DataPreprocessor()
#model = built_cnn()
model = loader.load_model_from_json('3_conv_adam_1e-4_cat_cross.json',
'3_conv_adam_1e-4_cat_cross.h5')
train_generator, validation_generator = preprocessor.get_image_generators()
history = train_network(model, train_generator, validation_generator)
loader.save_model(model, '3_conv_adam_1e-4_cat_cross.json',
'3_conv_adam_1e-4_cat_cross.h5')
#plot.plot_model(model, 'cnn_3_conv_RMSProp_1e-4.png')
plot.plot_history(history)
def train_network(model, train_generator, validation_generator):
model.compile(
loss='categorical_crossentropy',
optimizer=Nadam(lr=1e-4),
metrics=["accuracy"]
)
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples / train_generator.batch_size,
epochs=30,
validation_data=validation_generator,
validation_steps=validation_generator.samples / validation_generator.batch_size,
verbose=2
)
return history
if __name__ == '__main__':
preprocessor = DataPreprocessor.DataPreprocessor()
model = built_cnn()
#model = loader.load_model_from_json('3_conv_adam_1e-3_cat_cross_v2.json', '3_conv_adam_1e-3_cat_cross_v2.h5')
train_generator, validation_generator = preprocessor.get_image_generators()
history = train_network(model, train_generator, validation_generator)
loader.save_model(model, '3_conv_nadam_1e-4_v2.json', '3_conv_nadam_1e-4_v2.h5')
#plot.plot_model(model, 'cnn_3_conv_RMSProp_1e-4.png')
plot.plot_history(history)