def read_train_images():
train_data_path = os.path.join(DATA_PATH, 'train')
images = list_images(train_data_path)
total = int(len(images) / 2)
patient_classes = np.ndarray(total, dtype=np.uint8)
imgs = np.ndarray((total, IMG_ORIG_ROWS, IMG_ORIG_COLS), dtype=np.uint8)
imgs_mask = np.ndarray((total, IMG_ORIG_ROWS, IMG_ORIG_COLS), dtype=np.uint8)
print('Loading training images...')
i = 0
for image_path in images:
if 'mask' in image_path:
continue
image_name = os.path.basename(image_path)
name = image_name.split('.')[0]
patient_classes[i] = int(name.split('_')[0])
image_mask_name = name + '_mask.tif'
imgs[i] = load_img(os.path.join(train_data_path, image_name), grayscale=True)
imgs_mask[i] = load_img(os.path.join(train_data_path, image_mask_name), grayscale=True)
if i % 100 == 0:
print(f'Done: {i}/{total} images')
i += 1
return patient_classes, imgs, imgs_mask
def __getitem__(self, index):
tar_index = index % self.tar_size
clean = torch.from_numpy(
np.float32(load_img(self.clean_filenames[tar_index])))
noisy = torch.from_numpy(
np.float32(load_img(self.noisy_filenames[tar_index])))
clean_filename = os.path.split(self.clean_filenames[tar_index])[-1]
noisy_filename = os.path.split(self.noisy_filenames[tar_index])[-1]
clean = clean.permute(2, 0, 1)
noisy = noisy.permute(2, 0, 1)
return clean, noisy, clean_filename, noisy_filename
def __getitem__(self, index):
tar_index = index % self.rgb_size
clean_filename = os.path.split(self.clean_filenames[tar_index])[-1]
## Load Images
clean = np.float32(load_img(self.clean_filenames[tar_index]))
noisy = np.float32(load_img(self.noisy_filenames[tar_index]))
noisy = addWB(noisy, self.p)
if self.transform is not None:
clean, noisy = self.transform([clean, noisy])
return clean, noisy, clean_filename
def create_test_data():
train_data_path = os.path.join(DATA_PATH, 'test')
images = os.listdir(train_data_path)
total = len(images)
imgs = np.ndarray((total, 1, IMG_ORIG_ROWS, IMG_ORIG_COLS), dtype=np.uint8)
imgs_id = np.ndarray((total, ), dtype=np.int32)
print('Creating test images...')
i = 0
for image_path in images:
image_name = os.path.basename(image_path)
img_id = int(image_name.split('.')[0])
img = load_img(os.path.join(train_data_path, image_name), grayscale=True)
img = np.array([img])
imgs[i] = img
imgs_id[i] = img_id
if i % 100 == 0:
print(f'Done: {i}/{total} images')
i += 1
# Build all data set
print('Saving test samples...')
imgs = imgs[np.argsort(imgs_id)]
np.save(os.path.join(DATA_PATH, 'imgs_test.npy'), imgs)
print('Saving to .npy files done.')
def __getitem__(self, index):
tar_index = index % self.rgb_size
clean_filename = os.path.split(self.clean_filenames[tar_index])[-1]
noisy_filename = os.path.split(self.noisy_filenames[tar_index])[-1]
clean_filename = re.sub(GT, "", clean_filename, flags=re.I)
noisy_filename = re.sub(NOISY, "", noisy_filename, flags=re.I)
## Load Images
clean = np.float32(load_img(self.clean_filenames[tar_index]))
noisy = np.float32(load_img(self.noisy_filenames[tar_index]))
if self.transform is not None:
clean, noisy = self.transform([clean, noisy])
return clean, noisy, clean_filename, noisy_filename
def __getitem__(self, index):
tar_index = index % self.tar_size
LR = torch.from_numpy(
np.float32(load_img(self.LR_filenames[tar_index])))
LR_filename = os.path.split(self.LR_filenames[tar_index])[-1]
LR = LR.permute(2, 0, 1)
return LR, LR_filename
def __getitem__(self, index):
tar_index = index % self.tar_size
target = np.float32(load_img(self.target_filenames[tar_index]))
target = divisible_by(target, 16)
tar_filename = os.path.split(self.target_filenames[tar_index])[-1]
if self.transform:
target = self.transform(target)
target = torch.Tensor(target)
target = target.permute(2, 0, 1)
target = F.pad(target.unsqueeze(0),
(self.pad, self.pad, self.pad, self.pad),
mode='reflect')
target = self.blur(target).squeeze(0)
return target, tar_filename
def __getitem__(self, index):
tar_index = index % self.tar_size
target = np.float32(load_img(self.target_filenames[tar_index]))
target = divisible_by(target, 16)
tar_filename = os.path.split(self.target_filenames[tar_index])[-1]
target = torch.Tensor(target)
target = target.permute(2, 0, 1)
target = F.pad(target.unsqueeze(0),
(self.pad, self.pad, self.pad, self.pad),
mode='reflect')
target = self.blur(target).squeeze(0)
padh = (MAX_SIZE - target.shape[1]) // 2
padw = (MAX_SIZE - target.shape[2]) // 2
target = F.pad(target.unsqueeze(0), (padw, padw, padh, padh),
mode='constant').squeeze(0)
return target, tar_filename, padh, padw