search

sniklaus / softmax-splatting

an implementation of softmax splatting for differentiable forward warping using PyTorch
469 stars 58 forks source link

problem with generating synthetic image #11

Closed hogeman2 closed 4 years ago

hogeman2 commented 4 years ago

Nice work!

I have some problem reproducing your work. In my case, output image color is empathized. (some epoch G value is higher than original G value and some epoch R value is higher than original one)

Do you have any idea?

sniklaus commented 4 years ago

Thank you for the kind words! Would you mind elaborating on what you are referring to by "G" and "R"?

hogeman2 commented 4 years ago

Thanks for your reply Sorry for lack of my explanation G, R means color channel. I mean the colors from input image is changed after image synthetic module(Gridnet). green tree changes more green or less green or something.

sniklaus commented 4 years ago

I am afraid that I do not have enough information to help you debug this. This is also not an issue with softmax splatting per se, you can further see that no such issues were raised in #5.

hogeman2 commented 4 years ago

Thanks for your fast reply! I will close this issue.