A Study of RobustNet, a Domain Generalization Method for Semantic Segmentation

Domain Generalization alleviates the domain gap between training set and test set, improving performance of deep neural networks on out-of-dataset data. This opens possibility deploying models unlabelled data that were previously pretrained other datasets. In this article, we study ideas RobustNet [Choi et al. CVPR 2021], a recent method for in Urban-Scene Semantic Segmentation. Instead exposing network to wide range domains, tries separate domain-variant from domain-invariant features via whitening transformation. Then, only invariant are used training, which allows reduce time since no combination datasets is needed achieve invariance. addition, provide an easy-to-use demo where users can quickly their own compare results against state art semantic segmentation. **This MLBriefs source code has not been reviewed!** **The original [[available here|https://github.com/shachoi/RobustNet]] (last checked 2022/10/22).**