def __init__(self,
data_root_path,
folder,
crop_size=None,
trf_op=None,
Dataset_name="test"):
self.Dataset_name = Dataset_name
self.trf_op = trf_op
self.crop_size = crop_size
self.fix_list = [
RandomCrop(self.crop_size),
Normalize(),
Any2One(),
ToTensor()
]
if Dataset_name is "train":
self.image_paths = [
findFiles(data_root_path + "/{}/{}/*.IMA".format(x, y))
for x in folder["patients"] for y in folder["SliceThickness"]
]
self.image_paths = [x for j in self.image_paths for x in j]
else:
self.image_paths = findFiles("{}/{}/{}/*.IMA".format(
data_root_path, folder["patients"], folder["SliceThickness"]))
def __getitem__(self, idx):
full_image, quarter_image = self.imgset[idx]
if self.patch_size != 64:
crop_point = [0, 0]
else:
crop_point = np.random.randint(
512, size=2) ## Random interception patch
fix_list = [
Scale2Gen(scale_type="image"),
Normalize(normalize_type="image"),
RandomCrop(self.patch_size, crop_point),
ToTensor()
]
transf = self.Cal_transform(self.Dataset_name, self.pre_trans_img,
fix_list)
full_image = self.preProcess(full_image,
transf,
patch_size=self.patch_size)
quarter_image = self.preProcess(quarter_image,
transf,
patch_size=self.patch_size)
sample = {
"full_image": full_image.unsqueeze_(0),
"quarter_image": quarter_image.unsqueeze_(0)
}
return sample