Kaibo Tang
Kaibo Tang
I have tried the code suggested in #358, and there has been no problem for me. Hope it helps.
> Thanks for the clarification @ahoopes. So, if I were to have one category with labels A, B, C, another category with D, E, F and a number like age,...
Hi @Fjr9516, I don't have the exact same issue but I find it weird that the atlas generation model actually adds the rough template to the decoded atlas at the...
Hi @Fjr9516, I think I'm doing the same task as yours except that I am training from scratch on my own data. I have run 200 epochs so far. The...
> @kvttt just so i understand, the last image you showed is the final template when trained from scrach? That certainly seems much blurrier than I would expect. Yes. And...
> Hi @kvttt and @adalca, thanks for your replies! I think you are all right, the load_weights() function didn't work properly. It's caused by different settings between the trained model...
Hi @Fjr9516, similar to what you mentioned in your comment, I also compared using normalized and non-normalized age attributes. And in my case, the normalized version converges faster. Attached is...
Hi @Fjr9516, I wonder if you have figured out what configuration the pre-trained (atlas_creation_cond_NCC_1022.h5) model is using. It would be really helpful if you can give me some insight on...
Your edited NCC loss works for me. Thank you!
Hi M, Thanks for your updated NCC loss. I was just going over your code and I am having some questions about the following line: `cross = torch.clamp(cross, min =...