Open-VLFE (Open-set Mixed Domain Adaptation via Visual-Linguistic Focal Evolving)

The official Pytorch implementation of our paper ``Open-set Mixed Domain Adaptation via Visual-Linguistic Focal Evolving'', TCSVT 2025.

Open-set Mixed Domain Adaptation via Visual-Linguistic Focal Evolving

Bangzhen Liu, Yangyang Xu, Cheng Xu, Xuemiao Xu, and Shengfeng He

Paper

Environment

Python 3.9, Pytorch 2.0.1, TorchVision 0.15.2, Pytorch-lightning 2.3.3, clip 1.0. We used the pytorch-lightning library for memory efficient float16 training.

Data Preparation

Datasets

The dataset we adopted in our experiments includes: Office Dataset, OfficeHome Dataset, VisDA

Prepare dataset in data directory.

./data/Office31/amazon/images/ ## Office31
./data/OfficeHome/Real ## OfficeHome
./data/VisDA/Classification/train ## VisDA synthetic images
./data/VisDA/Classification/validation ## VisDA real images

File split

The file list for the datasets above are provided in ./txt folder, e.g.

./txt/Office31/source_amazon_obda.txt ## Office
./txt/OfficeHome/source_Real_obda.txt ## OfficeHome

The process scripts are provided in the "./utils/build_txt" folder, feel free to customize for any setting or new dataset by yourself.

Training

python train.py --expname <yourname> --dataset 'OfficeHome' 
            --source './txt/OfficeHome/source_Art_obda.txt'
            --target './txt/OfficeHome/target_cpr_obda.txt' 
            --network 'resnet50' --num_class 25 
            --batch_size 64 --num_workers 4 
            --is_train --epoch 40 
            --lr_f 0.001 --lr_n 0.01 --w_ent 0.1 --clip_thres 0.5 
            --gpu 0

Citation

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

@article{liu2025open,
title={Open-set mixed domain adaptation via visual-linguistic focal evolving},
author={Liu, Bangzhen and Xu, Yangyang and Xu, Cheng and Xu, Xuemiao and He, Shengfeng},
journal={IEEE Transactions on Circuits and Systems for Video Technology},
year={2025},
publisher={IEEE}
}

License

This repository is under MIT license.