xinntao / ESRGAN

ECCV18 Workshops - Enhanced SRGAN. Champion PIRM Challenge on Perceptual Super-Resolution. The training codes are in BasicSR.
https://github.com/xinntao/BasicSR
Apache License 2.0
5.91k stars 1.05k forks source link

Got different perceptual index result from the paper #74

Closed zpkosmos closed 4 years ago

zpkosmos commented 4 years ago

I use the https://github.com/roimehrez/PIRM2018 to test the result, but I got Your perceptual score is: 1.9127 Your RMSE is: 24.5861 for baboon from Set14 which in your paper the perceptual score is 1.98 and Your perceptual score is: 3.9012 Your RMSE is: 7.6694 for face from Set14 which in your paper the perceptual score is 3.64

I also test RCAN the perceptual score is 6.838 for face from Set14 which in your paper the perceptual score is 6.89

Was the model you provide is different from the model used in paper? or the perceptual score calculation is different

zpkosmos commented 4 years ago

Sorry, I find the answer in https://github.com/xinntao/ESRGAN/blob/24eca1d826ba7d059d86fdad60e2b04e0390419c/QA.md but the result of RCAN is still different from paper