[PTQ4SR] Unfair comparison in paper?

huawei-noah / Efficient-Computing

Efficient computing methods developed by Huawei Noah's Ark Lab

1.19k stars 208 forks source link

[PTQ4SR] Unfair comparison in paper? #53

Open Cheeun opened 1 year ago

Cheeun commented 1 year ago

In the CVPR2023 paper, "Toward Accurate Post-Training Quantization for Image Super Resolution", Table 5 (Sec. 4.2) of the paper seems unfair.

PAMS and FQSR are implemented on EDSR-baseline model (layers=16, dimension size=64),
while PTQ4SR (this works) is implemented on EDSR model (layers=32, dimension size=256).

Since the accuracy of EDSR and EDSR-baseline is largely different, isn't it unfair to compare methods with different backbones?

Are there comparisons made using the same EDSR-baseline backbone?

Looking forward to your code release!

Jeozhao commented 1 year ago

I found that this code you mentioned is no longer available. Could you please share a code with me? My email is 751350322@qq.com

Cheeun commented 1 year ago

Hi, unfortunately, I couldn't find the code as well, but I just left the issue here as the link from the paper leads to this repository.

tuzhijun commented 1 year ago

Sorry for that the code is not quite ready yet for personal resason. We will release our code within one month. And we will add QAT experiments based on EDSR (layers=16, dimension size=64) and present them in the latest code.

Angiemaster commented 1 year ago

Did the code release?

leenas233 commented 10 months ago

Sorry for that the code is not quite ready yet for personal resason. We will release our code within one month. And we will add QAT experiments based on EDSR (layers=16, dimension size=64) and present them in the latest code.

Will this code be released?