Compare two pred set and save their difference. Notice that path of two pred set should be set in this function. Parameter: nothing.

Compute the mean and std of dataset images. Parameter: dataset_name(str): name of the specified dataset. Return: means(list)

Crop dataset images. Crop image from :math:`pre_size` × :math:`pre_size` to :math:`after_size` × :math:`after_size` with :math:`overlap_size`

Crop whole large image. Crop the whole large image into :math:`crop_size`×:math:`crop_size` image without overlap. Notice source image path

Delete monochrome images in dataset. Delete whole black and whole white image in label directory and corresponding image in t1 and t2 direct

Display dataset image in wandb to inspect images. Notice that if mode is None, image path should be organized as :obj:`dataset_name`/`t1` or

u: r(B D L) delta: r(B D L) A: r(D N) B: r(B N L) C: r(B N L) D: r(D) z: r(B D L) delta_bias: r(D), fp32 ign

u: r(B D L) delta: r(B D L) A: r(D N) B: r(B N L) C: r(B N L) D: r(D) z: r(B D L) delta_bias: r(D), fp32 ign

Shuffle dataset images. Shuffle images in dataset to random split images to train, val and test later. Notice that image path should be org

Random sample specified ratio or number of dataset. Notice that if mode is None, image path should be organized as :obj:`dataset_name`/`t1`

Split dataset images. Split images to trian/val/test dataset with 7:2:1 ratio or corresponding specified number. Notice that image path sho

Verify correspondence between train/val/test dataset. Make sure there are corresponding images with the same name in :obj:`t1_images_dir`, :