Hotown / InCo_DA

Author's implementation in Pytorch for InCo (ECML-PKDD, 2022)

InCo: Intermediate Prototype Contrast for Unsupervised Domain Adaptation

This repository contains the author's implementation in Pytorch for InCo(Intermediate Prototype Contrast for Unsupervised Domain Adaptation).

Requirements

conda env create -f environment.yml
pip install -r requirements.txt

Training

• Download or soft-link the dataset under data folder (Supported datasets are ImageCLEF, Office-31, Office-Home)
• Run following command to train the InCo:

python inco/run.py --config config/${dataset}$/${task}$.json
python inco/run.py --config config/office/A-D-all.json

Acknowledgement

This code is built on PCS and TLlib. Thanks to the authors for sharing their codes.