search

DeokyunKim / Progressive-Face-Super-Resolution

Official Pytorch Implementation of Progressive Face Super-Resolution (BMVC 2019 Accepted)
260 stars 60 forks source link

PSNR #10

Open shuangliumax opened 4 years ago

shuangliumax commented 4 years ago

hello,Why don't you compare the performance PSNR of 2x and 4x?

DeokyunKim commented 4 years ago

The purpose of the research is x8 upscaling. Therefore, we did not compare and PSNR of 2x and 4x with baseline researches.

shuangliumax commented 4 years ago

@DeokyunKim But if x2,x4 is not good, will x8 be good?

DeokyunKim commented 4 years ago

The results of the last iteration in each step were experimentally good. Probably, if the results of x2 and x4 are not good, then x8 is also not good.

shuangliumax commented 4 years ago

@DeokyunKim hello,I would like to know if you have used your dataset to train and test other people's networks when comparing them to other people's methods, or if you have used your testset to test other people's networks directly

DeokyunKim commented 4 years ago

@DeokyunKim hello,I would like to know if you have used your dataset to train and test other people's networks when comparing them to other people's methods, or if you have used your test set to test other people's networks directly

We utilize pre-trained FSRNet and FSRGAN with the different datasets to compare with our methods, and we train the VSDR with the same data set with our trained model. We mentioned that in the paper.

yangyingni commented 4 years ago

I trained other people's network with my datasets.

------------------ 原始邮件 ------------------ 发件人: "notifications"<notifications@github.com>; 发送时间: 2020年1月9日(星期四) 晚上9:34 收件人: "DeokyunKim/Progressive-Face-Super-Resolution"<Progressive-Face-Super-Resolution@noreply.github.com>; 抄送: "泱渶"<2676426668@qq.com>; "Comment"<comment@noreply.github.com>; 主题: Re: [DeokyunKim/Progressive-Face-Super-Resolution] PSNR (#10)

@DeokyunKim hello,I would like to know if you have used your dataset to train and test other people's networks when comparing them to other people's methods, or if you have used your test set to test other people's networks directly

We utilize pre-trained FSRNet and FSRGAN with the different datasets to compare with our methods, and we train the VSDR with the same data set with our trained model. We mentioned that in the paper.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.