Adversarial Bayesian Augmentation (ABA)

Code for our paper: "Adversarial Bayesian Augmentation for Single-Source Domain Generalization" (ICCV 2023). To reproduce results for each benchmark, the following steps should be followed.

Train and Evaluate

• Digits: bash run_bal_mnist.sh

Acknowledgements

Part of the code structure is borrowed from ALT https://github.com/tejas-gokhale/ALT, BCNN https://github.com/kumar-shridhar/PyTorch-BayesianCNN/, RandConv https://github.com/wildphoton/RandConv and Sagnet https://github.com/hyeonseobnam/sagnet . We thank the authors of these papers.