bupt-ai-cz / iast-eccv2020 Goto Github PK

As far as I know, the size of the gta5 dataset is 24966. However, there are 24941 image-label pairs. The missing image ids are [ 2154, 2155, 2156, 2157, 2158, 4634, 4635, 4636, 4637, 7115, 7116, 7117, 7118, 7119, 9683, 9684, 9685, 9686, 9687, 9688, 12050, 12051, 12052, 12053, 12054]. Is there any reason? I appreciate any help you can provide.

Thanks for your code. Why p(c|x, w)/θ is used instead of only p(c|x, w), what`s the role played by θ？ And I can not find the corresponding code in this repo. I am unfamiliar with cross-domain semantic segmentation, so please forgive me if the question is too simple.