64 X 64 images

[koren12122]

There is a way to apply this method on Tiny-ImageNet (64 X 64)?
there is an existing model for that, or a new training is necessary?

[wyhuai]

It is best to have a 64 X 64 pertained model. If not, you can upsampling the 64x64 images into 256x256, then use DDNM for restoration, and then downsampling back to 64x64.

[koren12122]

thanks for the quick response.
Is there a way to upsample the image by using the DDNM? (The image is Out Of Distribution)

[Bulinglife]

Hi, how do you keep the image pixels the same after the training is done