Some Question About Experiment

[zqp1226358]

Hello, thank you for your reply. Currently I run the code but the result is not what I expected.

I suspect that there is still a problem with my preprocessing？ the coarse_generation is as follows

the fine_generation is as follows

the pseudo label is as follows, I pass in the source-like image to the pretrained model you provided

In addition, The "resnet_pretrain.pth" is missing, Is it convenient for you to provide it?

Looking forward to your reply. Thank you very much.

[puyln]

Hi, sorry to bother you, I want to ask you a question about "inversed" data and "pseudo" data. What images are contained in these folders?

[zqp1226358]

Hi，@puyln "inversed" data is fine_generation，"pseudo" data is generated by source-like image through pretained source model。

[qwerasdzxcvb]

Hello, I am reproducing your code and I am facing issues in Generation Stage： Every coarse generate doesn't show any content. I could not find a way to solve this issue, can you please provide some help？ Thanks for your time.

[zqp1226358]

@qwerasdzxcvb you can look this, in code try to initialize the input noise with image content. https://github.com/CityU-AIM-Group/SFDA-FSM/issues/2#issuecomment-1352951661

[qwerasdzxcvb]

@zqp1226358 Thank you for your reply, but the issue still persists after changing to content image. Is it convenient to leave a contact information? I have some specific questions I'd like to ask you.

[zqp1226358]

@qwerasdzxcvb You can contact me by google email

[cyang-cityu]

@zqp1226358 Thanks for your answers. @qwerasdzxcvb You'd better initialize the input noise image with original image as suggested by #2.

[cyang-cityu]

@zqp1226358 This is the resnet pretrained model. You can also apply our pretrained source model for coarse generation.