Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

About the place of pretrained CycleGAN #9

Open
IItaly opened this issue Mar 25, 2021 · 5 comments
Open

About the place of pretrained CycleGAN #9

IItaly opened this issue Mar 25, 2021 · 5 comments

Comments

@IItaly
Copy link

IItaly commented Mar 25, 2021

Thanks for your work.
I noticed that there are 18 pretrained model to make the dataset.
And the script to download the models have failed. So I want to know if I should put them to the "autogan/scripts" and then run the "run_test.py"?
I got the error when I try to "python run_test.py --dataset Cyclegan" :
NameErroe: "parser" is not defined

@IItaly
Copy link
Author

IItaly commented Mar 25, 2021

And How can I get the frequency data to train the classifier ?just like the pics referenced in your paper

@spongezhang
Copy link
Collaborator

Hi IItaly,

Sorry for the delayed reply. I left Columbia University and have less time to maintain this package.

For your first question, I'm not fully sure what models you are referring. There are two types of models,

  1. autogan model(generate autoencoder like images) the models are still here https://drive.google.com/file/d/1IZZ5QCLbzAmrPDHh7EdOigPCdz46zPc6/view.
    I mentioned this "Remove it once training is done. Once training is done, change all the autogan models' name in ./autogan/checkpoints/{category}_auto/latest_net_G_A.pth to ./autogan/checkpoints/{category}_auto/latest_net_G.pth" in the repo.
  2. the Classifier. The classifiers are missing, but training the classifiers with autogan/cyclegan images are relatively easy.

For the spectrum, here is how I calculate that.
https://github.com/ColumbiaDVMM/AutoGAN/blob/master/code/GAN_Detection_Train.py#L204

@IItaly
Copy link
Author

IItaly commented Apr 6, 2021

Thanks for your reply.Does your spectrum classifier use the color spectrum image shown in your paper? I tried to run your code for spectrum generation(just you have quoted ), and it generated a pure black image. Is it related to the args.mode parameter? I found that the parameter defaults to 0.

@spongezhang
Copy link
Collaborator

Haha, the color map is just the heatmap for the grayscale spectrum for visualization. Three channels are processed separately as shown in the code. :)

@IItaly
Copy link
Author

IItaly commented Apr 6, 2021

Haha, the color map is just the heatmap for the grayscale spectrum for visualization. Three channels are processed separately as shown in the code. :)

I get it.I was very confused by the color map, because I got the grayscale spectrum. Thank you for your answers.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants