def predictGenerator(test_path,
batch_size=2,
percent=1,
dim=(256, 256),
n_channels=1,
save_path='test_result/'):
if not os.path.exists(save_path):
os.makedirs(save_path)
files = random.sample(test_path, int(np.ceil(len(test_path) * percent)))
for i, item in enumerate(files):
img_array = img_load(os.path.join(item[0]), shape=dim, norm=True)
lab_array = lab_load(os.path.join(item[1]),
shape=dim,
norm=False,
binary=True)
img_save_name = str(i) + "_source_img_" + os.path.splitext(
os.path.split(item[0])[1])[0] + '.png'
lab_save_name = str(i) + "_source_label_" + os.path.split(item[1])[1]
imgdata = img_array * 255
labdata = lab_array * 255
io.imsave(os.path.join(save_path, img_save_name),
imgdata.clip(0, 255, imgdata).astype(np.uint8))
io.imsave(os.path.join(save_path, lab_save_name),
labdata.clip(0, 255, labdata).astype(np.uint8))
img_array = np.reshape(img_array, img_array.shape + (1, ))
img_array = np.reshape(img_array, (1, ) + img_array.shape)
yield img_array
def testGenerator(test_path, num_image=50, target_size=(256, 256), result_path='test_result/'):
filelist = travel_testfiles(test_path)
files = random.sample(filelist, num_image)
for i, item in enumerate(files):
img_array = img_load(os.path.join(test_path, item), shape=target_size, norm=True)
io.imsave(os.path.join(result_path, "%d_source.png" % i), img_array)
img_array = np.reshape(img_array, img_array.shape + (1,))
img_array = np.reshape(img_array, (1,) + img_array.shape)
yield img_array
def __data_generation(self, list_IDs_temp):
X = np.empty((self.batch_size, *self.dim, self.n_channels))
# print('X.shape', X.shape)
y1 = np.empty((self.batch_size, *self.dim, 1))
# print('y1.shape', y1.shape)
for i, item in enumerate(list_IDs_temp):
img_array = img_load(item[0], shape=(256, 256), norm=True)
lab_array = lab_load(item[1],
shape=(256, 256),
norm=False,
binary=True)
img_array = img_array.reshape(*img_array.shape, self.n_channels)
lab_array = lab_array.reshape(*lab_array.shape, self.n_channels)
if self.datagen:
seed = random.randint(1, 10000)
img_array = self.datagen.random_transform(img_array, seed)
lab_array = self.datagen.random_transform(lab_array, seed)
# print('each_img_shape', img_array.shape)
# print('each_lab_shape', lab_array.shape)
X[i, ] = img_array
y1[i, ] = lab_array
return X, y1