You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
Could I ask about the evaluation of the results of VQGAN. I generate the results by, python -u vqfr/test.py -opt options/test/VQGAN/test_vqgan_v1.yml
Then I evaluated the metrics by the commands as the figure below, but the results are lower than the results of VQFR provided in the paper. Since the VQGAN is an autoencoder of the HR images, is it correct that the results of VQGAN are lower than VQFR? Or do you know where did I evaluated wrong?
Thanks in advance!
The text was updated successfully, but these errors were encountered:
Sorry, I found the mistake, and my evaluation of the VQGAN is shown below. Am I correct since there are some metrics are still worse than stage 2?
Thanks!
Hi,
Could I ask about the evaluation of the results of VQGAN. I generate the results by,
python -u vqfr/test.py -opt options/test/VQGAN/test_vqgan_v1.yml
Then I evaluated the metrics by the commands as the figure below, but the results are lower than the results of VQFR provided in the paper. Since the VQGAN is an autoencoder of the HR images, is it correct that the results of VQGAN are lower than VQFR? Or do you know where did I evaluated wrong?

Thanks in advance!
The text was updated successfully, but these errors were encountered: