search

Ysz2022 / NeRCo

[ICCV 2023] Implicit Neural Representation for Cooperative Low-light Image Enhancement
https://openaccess.thecvf.com/content/ICCV2023/html/Yang_Implicit_Neural_Representation_for_Cooperative_Low-light_Image_Enhancement_ICCV_2023_paper.html
217 stars 15 forks source link

A question on LSRW dataset #15

Open wangsen99 opened 6 months ago

wangsen99 commented 6 months ago

Hello, thank you for the great work on this topic. I found your paper said: `` We train and test the model on LSRW dataset, which contains 1000 low-light-normal-light image pairs for training and 50 pairs for evaluation. '' But the LSRW I downloaded has about 5000+ image pairs for training including Huawei and Nikon. I'm so confused, hope to receive your reply.

Dicardo commented 6 months ago

Thank you very much for your interest in our work. In our study, while the LSRW dataset indeed contains over 5000+ image pairs, including both Huawei and Nikon images, we intentionally selected a subset of 1000 image pairs for our training set due to our limited hardware devices. For evaluation purposes, we used the entire set of 50 image pairs as specified in the paper to ensure that our results are comparable across different methods and to maintain a standard benchmark for evaluation. We believe that the performance could be further improved if we had the opportunity to train with the complete dataset.

BTW, if you find our work helpful, we would be honored if you consider citing it :)

wangsen99 commented 6 months ago

Thank you for the kind response, and it solved my confusion. Best wishes! ^_^