This repository hosts the implementation for our paper, "Domain Adversarial Neural Networks for Domain Generalization: When It Works and How to Improve". Our approach explores the application of domain adversarial neural networks for domain generalization.
To run the code, you first need to set up the environment. We use Conda for managing dependencies. The environment can be set up using the provided environment.yml file with the following command:
conda env create --file environment.ymlYou will also need to download the pretrained AlexNet model pretrained_alexnet.pth. You can find this file in the Files section at the following URL:
https://osf.io/87tjs/?view_only=99ca1354ca844d26be26516281ce964d
Once you have downloaded the file, place it in the src/models/caffenet directory.
To execute the experiments, navigate to the directory containing final_scripts and src. You can then run the bash scripts located in final_scripts/pacs and final_scripts/office_home.
We have included scripts for all of our main experiments in the paper. If you wish to run additional experiments, you can check the arguments that src.main accepts. This can be done either by looking at the code or using the help argument as follows:
python3 -m src.main --helpWe welcome any contributions to improve this project. Feel free to open issues or make pull requests.
This project is licensed under the MIT License. See the LICENSE file for more details.
If you find this work useful in your research, please consider citing our paper.
For any questions, please open an issue and we'll get back to you as soon as possible.