def train(self, dataset, batch_size, checkpoint_dir, steps_per_epoch,
nb_epocs, testing_ratio, validation_ratio, validation_steps):
'''
Starts training of the model with data provided by dataset.
'''
generator = self._get_data_generator(dataset, testing_ratio,
validation_ratio)
dataset_generator = generator.fcn_data_generator(
batch_size, self.patch_size, self.no_classes)
validation_generator = generator.fcn_data_generator(
batch_size, self.patch_size, self.no_classes, dataset='validation')
start_training(checkpoint_dir,
self.model,
dataset_generator,
steps_per_epoch,
nb_epocs,
callbacks=[],
validation_generator=validation_generator,
validation_steps=validation_steps)
def train(self, dataset, batch_size, checkpoint_dir, samples_per_epoc,
nb_epocs, testing_ratio, validation_ratio,
nb_validation_samples):
'''
Starts training of the model with data provided by dataset.
'''
patch_size = self.input_shape[1:3]
generator = self._get_data_generator(dataset, testing_ratio,
validation_ratio)
dataset_generator = generator.grid_dataset_generator(
batch_size, patch_size, self.grid_size)
validation_generator = generator.grid_dataset_generator(
batch_size, patch_size, self.grid_size, dataset='validation')
start_training(checkpoint_dir,
self.model,
dataset_generator,
samples_per_epoc,
nb_epocs,
callbacks=[],
validation_generator=validation_generator,
nb_val_samples=nb_validation_samples)
bb_out = Convolution2D(4, 1, 1, border_mode='same')(model)
model = Model(base_model.input, output=[class_out, bb_out])
optimizer = Adam(lr=1e-5, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
model.compile(optimizer=optimizer,
loss={'class_out': 'binary_crossentropy', 'bb_out': 'mean_squared_error'})
logger.info('Compiled fc with output:{}', model.output)
return model
if __name__ == '__main__':
dataset_dir = '/data/lrz/hm-cell-tracking/sequences_A549/annotations'
checkpoint_dir = '/data/training/yolo'
nb_objects = 5
batch_size = 100
samples_per_epoc = 5000
nb_epocs = 500
patch_size = (224, 224)
grid_size = (32, 32)
detector = YoloDetector(ResNet50, 'activation_48')
model = detector.build_model(nb_objects)
dataset = ImageDataset(dataset_dir)
dataset_generator = dataset.grid_patch_dataset_generator(batch_size, patch_size, grid_size=grid_size,
nb_objects=nb_objects)
start_training(dataset_dir, checkpoint_dir, model, dataset_generator, samples_per_epoc, nb_epocs)
dataset_dir = '/data/lrz/hm-cell-tracking/sequences_A549/annotations'
checkpoint_dir = '/data/training/cnn_conv3'
batch_size = 20
no_of_objects = 5
input_shape = (64, 64, 3)
samples_per_epoc = 80000
nb_epocs = 1000
model = cnn_conv3(input_shape=input_shape, out_size=no_of_objects)
image_dataset = ImageDataset(dataset_dir)
dataset_generator = image_dataset.patch_dataset_generator(
batch_size,
patch_size=input_shape[:2],
no_of_objects=no_of_objects,
dataset='training')
validation_generator = image_dataset.patch_dataset_generator(
batch_size,
patch_size=input_shape[:2],
no_of_objects=no_of_objects,
dataset='validation')
nb_val_samples = 100
start_training(dataset_dir,
checkpoint_dir,
model,
dataset_generator,
samples_per_epoc,
nb_epocs,
validation_generator=validation_generator,
nb_val_samples=nb_val_samples)