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

Differing FID Score #141

Open
AnqiGou opened this issue Dec 19, 2024 · 2 comments
Open

Differing FID Score #141

AnqiGou opened this issue Dec 19, 2024 · 2 comments

Comments

@AnqiGou
Copy link

AnqiGou commented Dec 19, 2024

I referred to the documentation for the measurement of the FID score, first I used the ckpt1.1 model to generate 1000 test images, and calculated the FID score with the laion-40k images in the FID folder, and the score was 57.86, which is very different from the 33.54 in the paper. The results of OCR and NED are not much different. What could be the reason for this?

@tyxsspa
Copy link
Owner

tyxsspa commented Dec 19, 2024

Hi,
Are you using eval/eval_fid.sh to calculate FID?
The generate 1000 images all in English right? (if they are in Chinese, dont compare FID with laion-40k)

@AnqiGou
Copy link
Author

AnqiGou commented Dec 22, 2024

Yes, all the 1000 images are in English (with test1k.json file), and I calculate FID with eval/eval_fid.sh. I have on idea why the fid score is so much worse than yours, are there any details that need to be paid attention to?

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