AlessioGalluccio
AlessioGalluccio
Hi @stvogel, I use torch 1.10.0 and I never got that error. Did you modify the code in some way? If there is a way to solve this issue, I'm...
That's a great idea. I updated it. Thank you very much. Best, Alessio
Repo is complete when we get the results comparable to the ones of the paper! Best, Alessio
Hello @xfby2016, I haven't tested the mvtec dataset yet. I'm still trying to recreate the model of the paper. For example, I still have issues in understanding which type of...
I haven't understood it yet. I'm open to suggestions
Hi Emilie! I haven't managed to reproduce them yet, but others had some progress, I'm currently working on my Master thesis and, since I couldn't reproduce the results of the...
Hi @mjack3, I'm really glad to find some help in this project. Thank you very much for your proposal, I accept. This paper is quite obscure. The problem you are...
@mjack3 I tried to contact Yushuang Wu through a university e-mail I found, but I got no answer. I haven't found the e-mail of the other authors
The permutation of channels must be fixed during training. As @gathierry mentioned, it's necessary for normalizing flows. > @AlessioGalluccio just a small remark: For anomaly score calculation (global and pixelwise)...
Looking a CFlow AD, it does in utils.py `logp = C * _GCONST_ - 0.5*torch.sum(z**2, 1) + logdet_J` He computes the positive likelihood instead of the negative one. In fact,...