SC-CAM + RW

[pierrefournier752]

Hi, thanks a lot for sharing your very interesting work.

I am running the code to try to reproduce the results in the paper (still running the second round). If I understood correctly, the results that I should get after this are the ones shown in the Table (~50.9), which corresponds to the model without the RW in AffinityNet. I was wondering how I should proceed to run this model (SC-CAM + RW) to get the 63.4 mIOU reported in the paper.

Thanks!

Pierre

[Juliachang]

Hi @pierrefournier752,
Yes, your understanding is correct. I used the source code of AffinityNet to apply the random walk (RW). Please find it at their repo: https://github.com/jiwoon-ahn/psa.

[pierrefournier752]

Thanks for your quick answer.

So, to be on the same page:

• first, I train with the train_cls.py function on this repo.
• And after that I use the infer_aff.py on their repository?

Otherwise, which is the sequence of functions that I should use?

Thanks again.
Best

[TyroneLi]

Hi @pierrefournier752,
Yes, your understanding is correct. I used the source code of AffinityNet to apply the random walk (RW). Please find it at their repo: https://github.com/jiwoon-ahn/psa.

Do your use your infer cam to perform Train AffinityNet with the labels procedure?? Or how your perform random walk without training a AffinityNet?? Could you provide more details? Thanks

[Juliachang]

Hi @TyroneLi,

You can use the provided model from the AffinityNet to perform random walk and you should already obtain good results.

[Alice-Shen]

@TyroneLi Do you solve the problem inferring cam with random walk? I do not know how should I get the weight file.

[OpencvW]

Hi. I got all black image when I run infer_aff.py file. Does anyone have the same problem as me？please help me，thank you!