Self-Supervised Graph Neural Network for Multi-Source Domain Adaptation

This project is the official implementation of ``Self-Supervised Graph Neural Network for Multi-Source Domain Adaptation'' in PyTorch, which is accepted by ACM MM 2022.

<p align="center"> <img src="docs/img4.jpg" /> </p>

Prerequisites

• Python 3.6
• PyTorch 1.4.0
• CUDA 9.0 & cuDNN 7.0.5

Dataset Preparation

• Office-Caltech
• Office-31
• Office-Home
• DomainNet

Pre-trained Models

Training

To train the full model of SSG, simply run:

python train.py --use_target --save_model --target clipart \
                --checkpoint_dir $save_dir$

Like

python train.py --use_target --save_model --target clipart

A large body of the code is borrowed from "Learning to Combine: Knowledge Aggregation for Multi-Source Domain Adaptation". Thanks!

Citation

If this work helps your research, please cite the following paper:

@inproceedings{yuan2022self,
  title={Self-Supervised Graph Neural Network for Multi-Source Domain Adaptation},
  author={Yuan, Jin and Hou, Feng and Du, Yangzhou and Shi, Zhongchao and Geng, Xin and Fan, Jianping and Rui, Yong},
  booktitle={Proceedings of the 30th ACM international conference on multimedia},
  year={2022}
}