def train(self, options):
print('-' * 30)
print('Loading and preprocessing train data...')
print('-' * 30)
imgs_train, imgs_mask_train_list = load_train_data(options)
print('Creating and compiling model...')
model = self.get_unet()
for imgs_mask_train in imgs_mask_train_list:
if not imgs_mask_train.endswith('.npy'):
print('Ignored: ' + imgs_mask_train)
print('-' * 30)
continue
print('Loading mask: ' + imgs_mask_train)
model_name = imgs_mask_train.split('.')[0]
imgs_mask_train = np.load(
os.path.join(options.inputdir,
imgs_mask_train)).astype('float32')
imgs_train = imgs_train.astype('float32')
imgs_mask_train /= 255. # scale masks to [0, 1]
imgs_train /= 255. # scale masks to [0, 1]
print('-' * 30)
print('Training for', model_name)
weight_dir = options.outputdir + '/weights'
if not os.path.exists(weight_dir):
os.mkdir(weight_dir)
model_checkpoint = ModelCheckpoint(os.path.join(
weight_dir, model_name + '.h5'),
monitor='val_loss',
save_best_only=True)
log_dir = options.outputdir + '/logs'
if not os.path.exists(log_dir):
os.mkdir(log_dir)
csv_logger = CSVLogger(os.path.join(log_dir, model_name + '.txt'),
separator=',',
append=False)
print('Fitting model...')
print('-' * 30)
model.fit(imgs_train,
imgs_mask_train,
batch_size=1,
epochs=options.epochs,
verbose=1,
shuffle=True,
validation_split=0.10,
callbacks=[model_checkpoint, csv_logger])
print('Finished training: ', model_name)
print('-' * 30)
print('Training finished, total: ' + str(len(imgs_mask_train_list)) +
' models generated.')
print('-' * 30)
def train(self, options):
print('-' * 30)
print('Loading and preprocessing train data...')
print('-' * 30)
imgs_train, imgs_mask_train = load_train_data(options)
imgs_mask_train = imgs_mask_train.astype('float32')
imgs_train = imgs_train.astype('float32')
imgs_mask_train /= 255. # scale masks to [0, 1]
imgs_train /= 255. # scale masks to [0, 1]
print('-' * 30)
print('Creating and compiling model...')
print('-' * 30)
model = self.get_unet()
weight_dir = options.outputdir + '/weights'
if not os.path.exists(weight_dir):
os.mkdir(weight_dir)
model_checkpoint = ModelCheckpoint(os.path.join(
weight_dir, project_name + '.h5'),
monitor='val_loss',
save_best_only=True)
log_dir = options.outputdir + '/logs'
if not os.path.exists(log_dir):
os.mkdir(log_dir)
csv_logger = CSVLogger(os.path.join(log_dir, project_name + '.txt'),
separator=',',
append=False)
print('-' * 30)
print('Fitting model...')
print('-' * 30)
model.fit(imgs_train,
imgs_mask_train,
batch_size=1,
epochs=options.epochs,
verbose=1,
shuffle=True,
validation_split=0.10,
callbacks=[model_checkpoint, csv_logger])
print('-' * 30)
print('Training finished')
print('-' * 30)