search

limbee / NTIRE2017

Torch implementation of "Enhanced Deep Residual Networks for Single Image Super-Resolution"
652 stars 146 forks source link

about PSNR and SSIM on unknown track #24

Closed ZJiaQi closed 5 years ago

ZJiaQi commented 6 years ago

Hi~Thanks for your superior works. I want to know the PSNR and SSIM that you achvevd on the unknown track of DIV2K dataset. Thank you.

limbee commented 6 years ago

Sorry, we don't have any results on the unknown track in the paper. There's a detailed report on the NTIRE2017 Challenge (http://ieeexplore.ieee.org/abstract/document/8014883/) in which you can find our challenge scores, but models submitted to the challenge has slightly different specifications from those for the paper. However, you can make your own results using our code. Training options should be: -degrade unknown -augUnkDIV2K false. When testing, do not enable the -selfEnsemble option. Follow all the other steps as same as in the bicubic scenario. The final Matlab evaluation code will give you the PSNR and SSIM.

As soon as possible, we'll update our arxiv paper and README to include the results of unknown track.

ZJiaQi commented 6 years ago

Thanks for your kind reply.