Too high FID in quantitative analysis #58
Description
Hello, I used the edge2shoes dataset to reproduce the process, and calculated the FID of the 200 and ground_truth files in the generated sample_to_eval. The result was 45.977917507375594, which is much higher than the result in your paper. However, the LPIPS result is in line with expectations, which is 0.184. Which step did I go wrong?