search

DmitryUlyanov / deep-image-prior

Image restoration with neural networks but without learning.
https://dmitryulyanov.github.io/deep_image_prior
Other
7.79k stars 1.42k forks source link

About the input image, it seems not all image can be handled without errors #107

Open drrobincroft opened 1 year ago

drrobincroft commented 1 year ago

I have tried some my own images as input for DIP. It seems some images would lead to errors like Argument #6: Padding size should be less than the corresponding input dimension, but got: padding (1, 1) at dimension 2 of input 4 or The size of tensor a (3) must match the size of tensor b (4) at non-singleton dimension 1 I cannot tell the difference among the input images, except for contents.

Is there some preconditions for the input image? This is really wired.