Open LuongTuanAnh163002 opened 1 week ago
Hello~ We use the public FID evaluation code available at https://github.com/omni-us/research-GANwriting
Thank you so much, I used FID evaluation of https://github.com/omni-us/research-GANwriting also read through the code and found that the code is no different from the FID calculation codes that I tried before (pytorch_fid) and the result is still incorrect (FID: 40). The result I evaluated on the test set of IAM and the correct result in the paper should be FID: 15.73
I read the code and referred to the code (test.py file) and wrote an infer.py file myself to generate fake images on the IAM test set and compare it with the real image. However, the generated image I see has a large white space behind. Is that a problem? Example:
Hello author, I am not able to find the code to confirm the evaluation metric you stated in the paper(FID: 15.73) with your checkpoint. I have tried the FID evaluation codes available online but the results are not the same as in your paper(my FID is 39.15). Can you provide me the code and instructions with your checkpoint to get the exact results as you stated in the paper? Thank you so much!