https://github.com/xudonmao/VMT

https://github.com/xudonmao/VMT

Last synced: 3 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/xudonmao/VMT
• Owner: xudonmao
• License: mit
• Created: 2019-07-03T14:15:24.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-08-15T14:55:45.000Z (about 5 years ago)
• Last Synced: 2023-06-16T11:30:35.094Z (over 1 year ago)
• Language: Python
• Size: 18.6 KB
• Stars: 19
• Watchers: 1
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• Awesome-Mixup - [Code

README

        
# VMT

Code for [Virtual Mixup Training for Unsupervised Domain Adaptation](https://arxiv.org/abs/1905.04215).

### Acknowledgments

This code is based on [dirt-t](https://github.com/RuiShu/dirt-t).

### Dependencies

```python

numpy==1.14.1

scikit_image==0.13.1

scipy==1.0.0

tensorflow_gpu==1.6.0

tensorbayes==0.4.0

```

### Train

1. Run VMT

```

python -u run_dirtt.py --datadir data --src mnist --trg svhn --inorm 1 --run 0 --dirt 0 --dw 0.01 --svw 1 --tvw 0.06 --tcw 0.06 --smw 1 --tmw 0.06

```

2. Run DIRT-T 

```

python -u run_dirtt.py --datadir data --src mnist --trg svhn --inorm 1 --run 0 --dirt 5000 --dw 0.01 --svw 1 --tvw 0.06 --tcw 0.06 --smw 1 --tmw 0.06

```

### Citation

If you use this work in your research, please cite:

    @article{arxiv1905.04215,

      author = {Xudong Mao and Yun Ma and Zhenguo Yang and Yangbin Chen and Qing Li},

      title = {Virtual Mixup Training for Unsupervised Domain Adaptation},

      journal = {arXiv preprint arXiv:1905.04215},

      year = {2019}

    }