Official implementation for NOUN and PRONOUN

[TIP-2021] Adversarial Domain Adaptation with Prototype-Based Normalized Output Conditioner

1. Prepare the running environment.

Packages used in our anaconda virtual environment are listed in 'environment.yml'.

2. Prepare the data.

Download the office-home dataset. Generate the data list and replace the path 'xxxx' in 'data/office-home/*.txt' with your own path.

3. Run our demo code.

The executive command for each domain adaptation (DA) method is shown in 'demo_home_ac.sh'. Using this script, you can reproduce our results of all these DA methods on the 'A to C' task of Office-home DA benchmark.

Citation

If you find this code useful for your research, please cite our paper

@article{hu2021adversarial,
    title={Adversarial Domain Adaptation with Prototype-Based Normalized Output Conditioner},
    author={Hu, Dapeng and Liang, Jian and Hou, Qibin and Yan, Hanshu and Chen, Yunpeng},
    booktitle={IEEE Transactions on Image Processing (TIP)},
    pages={9359--9371},
    volume={30},
    year={2021},
    publisher={IEEE} }

Contact

• [email protected]
• [email protected]