# ignore warnings
warnings.filterwarnings("ignore")
# configuration
config = ImageConfig()
config.model_description = ''
config.time = ''
config.lr = 3e-4
config.operation = ConfigOpt.TRAIN
config.images_dir = '../data/golgi images 0'
config.masks_dir = '../data/golgi masks 0'
config.mean_map = 'mean_map_{}.pic'.format('')
# prepare data (augmentation)
if config.operation == ConfigOpt.AUGMENTATION:
augmentation(config)
if isinstance(config.operation, ConfigOpt) and \
config.operation is not ConfigOpt.AUGMENTATION:
# create model
model = Unet(config=config)
# load data
x, y = load_images_train_data(model, img_num=8000)
# date generator
date_gen = image_data_generator()
# run model
model.run_model(x, y, use_generator=False, date_gen=date_gen)
config.mean_map = os.path.join(root_path, 'code/results/mean_map',
'mean_map_{}.pic'.format(config.time))
# prepare data
# run
if config.operation is not ConfigOpt.AUGMENTATION:
# create model
model = Unet(config=config)
# load data
if config.operation is ConfigOpt.TRAIN:
x, y = load_images_train_data(model)
# data generator
data_gen = image_data_generator()
# run model
model.run_model(x, y, use_generator=False, data_gen=data_gen)
elif config.operation is ConfigOpt.PREDICT:
x = load_image_test_data(model, config.images_dir, 100)
y_gt = load_image_test_data(model, config.masks_dir, 100)
# data generator
data_gen = image_data_generator()
# run model
config.time = '117_229'
weights = os.path.join(config.root_path, 'code/results/weights', config._weights)
y = model.run_model(x, weights=weights)
print(y.shape)
for i in range(y.shape[0]):
cv2.imshow('predicted image',
np.concatenate((x[i][:,:,0], y[i][:,:,0], y_gt[i][:,:,0]), axis=1))
cv2.waitKey()
