Adversarially Learned Transformations (ALT)

Code for our paper: "Improving Diversity with Adversarially Learned Transformations for Domain Generalization" (WACV 2023). To reproduce results for each benchmark, the following steps should be followed.

Data Download:

First, download the data using the following instructions:

1. Digits -- data will download automatically if you run run_alt_mnist.sh
2. PACS -- download from https://mega.nz/#F!jBllFAaI!gOXRx97YHx-zorH5wvS6uw
3. OfficeHome can be downloaded from the official release https://www.hemanthdv.org/officeHomeDataset.html

Train and Evaluate

• Digits: bash run_alt_mnist.sh
• PACS: bash run_alt_pacs.sh
• Office-Home: bash run_alt_officehome.sh

Acknowledgements

Part of the code structure is borrowed from RandConv https://github.com/wildphoton/RandConv and Sagnet https://github.com/hyeonseobnam/sagnet . We thank the authors of these papers.