pvnieo/low-light-image-enhancement

Python implementation of two low-light image enhancement techniques via illumination map estimation

low-light-image-enhancement's Issues

What are the results of the two models?

Hello, I compared the results of the two models and found no change without changing the parameters g and l. Which model do you think is better?

Result differs from the dual illumination paper

I have tried running the dual algorithm and comparing it to the result from the paper, but they are different.

This is the original and enhanced version from the paper:

This is the output I get from running the dual script for this project:

I am using the default parameters, g=0.6 and l=0.15. The results seem more noisy and oversaturated than the results displayed in the paper, and the contrast is different.

Is there a combination of parameters that would give the same results as the paper that I can try, or is it possible that there is a bug in the algorithm? Thanks.

The Matlab-implementation of DUAL is much fast than this version. Can this be fixed with an alternative LinAlg Solver?

Why does it take so long to process

How can i train with my own datasets??

GPU support

Does the model run on a GPU? If not, how can this be done? And are multiple GPUs supported?

I also noticed that images larger than 1000px are skipped due to lack of memory. I get pretty slow performance on RTX3080