Closed Madeline-hyh closed 10 months ago
For fair comparisons, we follow the previous methods[1,2] to uniformly sample images/patches from the dataset for network training. You could refer to the image sampling code.
[1]All in one bad weather removal using architectural search, CVPR 2020 [2]Learning multiple adverse weather removal via two-stage knowledge learning and multi-contrastive regularization: Toward a unified model, CVPR 2022
Hello! This is a very good work. I am very interested in your work. I wonder if you can provide the sampled dataset, thank you~