def load_image(data, tar_size=(2048, 2048)):
image = prep.imread(data.image)
label = prep.imread(data.label, c=0)
vessel = prep.imread(data.vessel, c=0)
if image.shape[:2] != tar_size:
image = prep.resize_image(image, tar_size=tar_size)
if label.shape[:2] != tar_size:
label = prep.resize_image(label, tar_size=tar_size)
if vessel.shape[:2] != tar_size:
vessel = prep.resize_image(vessel, tar_size=tar_size)
return image, label, vessel
def _transform_label_data(file_path):
img = imread(file_path, c=0)
if img.max() > 1:
img = img.astype(np.float) / 255.0
label_img = np.ones(shape=(img.shape[0], img.shape[1], c.layer_num))
for i in range(c.layer_num):
label_img[:, :, i] = (img == i).astype(np.float)
return label_img
def clear_images():
image_dicts = libfi.getfiledicbyext(_c.model_image_dir, ext='jpg')
label_dicts = libfi.getfiledicbyext(_c.model_label_dir, ext='png')
for k, v in label_dicts.items():
label = prep.imread(v, 0)
if np.sum(label) <= 1:
os.remove(v)
os.remove(image_dicts[k])
print(k)
for k, v in image_dicts.items():
if k not in label_dicts:
os.remove(v)
def _transform_image_data(file_path):
img = imread(file_path)
img = img.astype(np.float) / 255.0
return img