rakhimovv / generativelatentflow Goto Github PK

What this script does is it calculates the FID score for some random images:

predictor = InceptionPredictor(output_dim=64)
true_images = torch.rand(32, 3, 299, 299)
art_images = torch.rand(32, 3, 299, 299)

inc_true = predictor(true_images).cpu().numpy()
inc_art = predictor(art_images).cpu().numpy()

inc_fid = frechet_distance(inc_true, inc_art)

If you are able to run this script with no errors, then your problem is probably related to the inputs you are feeding. What are yours true_images and art_images? How did you obtain those?

Originally posted by @fursovia in #1 (comment)

I have tried a few times the generative latent flow model for Cifar10 dataset, but I have always got this error after a few epochs, (Imaginary or complex number in FID score), which stems from the following line in fid_score.py. I would really appreciate it if you could let us know how to fix this issue.

raise ValueError('Imaginary component {}'.format(m))

It seems this code only compute the FID of CIFAR-10,CelebA', 'dots', I am wondering how to compute the fid of the mnist