FID evaluation

[LeeUichann]

hi, I have a question.
Is the FID score in Table 1 calculated using CleanFID between the image set generated by SD1.4 with prompts from the coco_30k_10k dataset and the image set generated by the unlearned model with the same prompts?
Could you provide a detailed explanation of the procedure used to calculate the FID in Table 1?
Thank you.

[dokko1]

Hi, Did you figure out this issue?

When I tried to reproduce the results using the coco_30k_10k datasets provided by the authors, I calculated the Clean-FID between images generated by the SD-v1.4 model & those generated with Safree applied. The result I obtained was 8.241.

[YtengTan]

Same doubts ...
And was the script for FID, CLIP and TIFA evaluations even included in this repository?
I couldn't find it.