Why are there big errors in my test results?
Opened this issue · 1 comments
yut94 commented
Hello. Thank you very much for your work. However, there are large errors in the results I obtained using the pre-trained model and code you provided.
The average PSNR I tested on NVIDIA 3090 + PyTorch 1.8.2 was 19.84.
I switched to NVIDIA V100 + PyTorch 1.8.2 or NVIDIA V100 + PyTorch 1.5.0, and the average PSNR was also 19.84.
I could never reach the 21.32 in the paper.
Why is this?
AndersonYong commented
Hi, sorry for the late reply, I guess your issue may be similar with this one (https://github.com/AndersonYong/URetinex-Net/issues/1). The reason is that results are affected by illumination adjust ratio.