Ysz2022 / NeRCo

[ICCV 2023] Implicit Neural Representation for Cooperative Low-light Image Enhancement

Home Page:https://openaccess.thecvf.com/content/ICCV2023/html/Yang_Implicit_Neural_Representation_for_Cooperative_Low-light_Image_Enhancement_ICCV_2023_paper.html

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

wrong function

lodew920 opened this issue · comments

Did the code on lines 85 to 88 of the mlp.py file not work? I found that some code is commented, are these functions commented incorrectly?
There is something wrong with the function “def query_rgb(self, inp, coord, cell=None)“?

Thanks for your interest in our work! :)

The commented part is the old version we have used in the previous experiments, we rewrite it with a simpler way, so the old version is commented. As far as I know, there is nothing wrong in “def query_rgb(self, inp, coord, cell=None)“. If you have any questions about it, please tell me and look forward to discussing with you!

Thank you very much for your prompt reply
I found that this code is not used in the following code?
Snipaste_2023-09-05_18-08-16
Besides, the subtraction of these two coordinates is equal to 0?
image
Looking forward to your reply, maybe I misunderstood

I'm very sorry😖, I fell ill a few days ago😰 and I only replied to your message now.

For your first question, admittedly, coord_ has not been used in our code. But we have used it in the previous version to get q_feat, as the commented part shows.

For your second question, our code is partly based on LIIF[1], a super-resolution method. In its code, "coord_" is the coordinate of high-resolution image and "q_coord" is the coordinate of low-resolution image, it uses "coord_" and "q_coord" to calculate the relative coordinate "rel_coord". Here as we do not need super-resolution, our "rel_coord" should be set as 0.

[1] Yinbo Chen, Sifei Liu, and Xiaolong Wang. Learning continuous image representation with local implicit image function. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2021.

Thank you very much for your answer. I am very interested in the structure of your code! Thanks again for your answer