search

ZhexinLiang / CLIP-LIT

[ICCV 2023, Oral] Iterative Prompt Learning for Unsupervised Backlit Image Enhancement
https://zhexinliang.github.io/CLIP_LIT_page/
274 stars 23 forks source link

Is there any special design for low light enhancement? #14

Closed order-a-lemonade closed 10 months ago

order-a-lemonade commented 11 months ago

can we just change the dataset for other image enhancement tasks?

ZhexinLiang commented 11 months ago

Hi @order-a-lemonade, thanks for you interest in our work.

There is no special design for low light enhancement (maybe using the PPM block can be a small design). I think for other image enhancement task, you can try more commonly-used network, like Restormer, because our network is relatively simple and maybe cannot handle more complex tasks. For example, you can add a deraining block for deraining tasks. Based on my experiments, directly using our method can do deraining but the performance is not SOTA.

Feel free to contact me if you have any other questions.

order-a-lemonade commented 10 months ago

Hi @order-a-lemonade, thanks for you interest in our work.

There is no special design for low light enhancement (maybe using the PPM block can be a small design). I think for other image enhancement task, you can try more commonly-used network, like Restormer, because our network is relatively simple and maybe cannot handle more complex tasks. For example, you can add a deraining block for deraining tasks. Based on my experiments, directly using our method can do deraining but the performance is not SOTA.

Feel free to contact me if you have any other questions.

An awesome work! and thanks for your reply.