你好，我最近打算做医学图像配准，刚开始做，还不太了解。请问你这个图像配准的评价指标是什么，训练的时候loss是什么。谢谢

Hi, I have to say this is not image registration but a image synthesis/generation model. Registration means you predict the spacial transformation of the image and transform the image using interpolation, Not that there's no mapping between image intensity but just spacial transformations.

else:
--> 124 img_new_value = slice[img_position]

Hi I am getting this error while deforming celebA image. I am opening the image in PIL and using convert('L') and passing numpy array to deforme function.
This is working for emoji but not for my current dataset.