Is the CoCo FID indicator abnormal? #130
Description
Hello author, I am trying to perform an FID test for COCO 30K using the code from the repository that tests FID, but the FID result I got when I ran the test was as high as 40.980.
I'm not sure where the problem is right now.
The ground truth path for fid_score.py is the path to the .npz compressed file obtained from T2IBenchmark.datasets.get_coco_fid_stats.
I wonder if anyone else has encountered the same problem as me.