Awesome

HDA implemented in PyTorch

code release for "Heuristic Domain Adaptation"(NIPS2020)

One sentence highlight

We address the construction of domain-invariant and domain-specific representations from the heuristic search perspective.

Poster

Enviroment

• pytorch = 1.3.0
• torchvision = 0.4.1
• numpy = 1.17.2
• pillow = 6.2.0
• python3.7
• cuda10

To install the required python packages, run

pip install -r requirements.txt

Dataset

Office-Home dataset can be found here.

Domainnet dataset can be found here.

The training of HDA could be utilized by changing the path of the dataset, such as the txt files in data/UDA_officehome/Art.txt.

Also, the txt files for SSDA and MSDA should be compressed.

cd data
unzip data.zip

Train

UDA on Office-Home

bash scripts/run_uda.sh

MSDA on Domainnet

bash scripts/run_msda.sh

SSDA on Domainnet

bash scripts/run_ssda.sh

Citation

If you use this code for your research, please consider citing:

@inproceedings{cui2020hda,
 author = {Cui, Shuhao and Jin, Xuan and Wang, Shuhui and He, Yuan and Huang, Qingming},
 booktitle = {Advances in Neural Information Processing Systems},
 pages = {7571--7583},
 publisher = {Curran Associates, Inc.},
 title = {Heuristic Domain Adaptation},
 volume = {33},
 year = {2020}
}

Contact

If you have any problem about our code, feel free to contact

• hassassin1621@gmail.com

or describe your problem in Issues.

Note

量子位

Supplemantary could be found in google driver and baidu cloud with 8yut.