search

cuiziteng / Illumination-Adaptive-Transformer

🌕 [BMVC 2022] You Only Need 90K Parameters to Adapt Light: A Light Weight Transformer for Image Enhancement and Exposure Correction. SOTA for low light enhancement, 0.004 seconds try this for pre-processing.
Apache License 2.0
480 stars 44 forks source link

About LOL dataset #6

Closed xiaobiaodu closed 2 years ago

xiaobiaodu commented 2 years ago

Hi

Could you please tell me the difference between the original LOL dataset with LOL-v2?

Best

cuiziteng commented 2 years ago

The original LOL has 15 test image, and the renewed LOL has 100 test image, their training image is also different, you can replace this dataset to LOL-v1, it's both OK.

xiaobiaodu commented 2 years ago

Thanks for your reply. Could you please tell me how you achieve RGB to RAW and ISP process? I didn't see that in your script.

cuiziteng commented 2 years ago

Thanks for your question, in this arxiv version we have not clearly show this process, this is also mentioned by the ECCV's reviewer, so we have renewed our arxiv version, the renewed arxiv version would be released in these days, please keep an eye on this project, thanks!

xiaobiaodu commented 2 years ago

Hi, I am still confused how you convert the original LOL dataset into renewed LOL. I want to ensure the results shown in your paper are fair. And your code does not contain about ISP and RGB to RAW process.

cuiziteng commented 2 years ago

We will renew the LOL-v1 results in the lastest version, and the RGB2RAW process is the network encoder (local branch), please see the paper if you confused.