ICON

Code release for "Make the U in UDA Matter: Invariant Consistency Learning for Unsupervised Domain Adaptation" (NeurIPS 2023). Paper is available here.

Prerequisites

• torch>=1.7.0
• torchvision
• qpsolvers
• numpy
• prettytable
• tqdm
• scikit-learn
• webcolors
• matplotlib

Training

Replace {data_dir} with the dataset directory. Missing datasets will be downloaded automatically. Replace {log_dir} with the logging directory (for storing model checkpoints, tensorboard logs and console logs). For Office-Home, source (-s) and target domain (-t) takes values from {'Ar', 'Cl', 'Rw', 'Pr'}.

VisDA-2017

CUDA_VISIBLE_DEVICES=0 python run_icon.py {data_dir} -d VisDA2017 -s Synthetic -t Real -a resnet50 --epochs 50 --lr 0.002 --per-class-eval --temperature 3.0 --center-crop --w-transfer 0.08 --w-st 1.0 --threshold 0.97 --log-root {log_dir} --batch-size 28 --optim sgd --con-start-epoch 5 --con-mode sim --w-inv 0.25 --inv-start-epoch 5 --back-cluster-start-epoch 9 --topk 3 --dim-reduction umap --reduced-dim 50 --eqinv --exp-name visda_reproduce --seed 0

Office Home

CUDA_VISIBLE_DEVICES=0 python run_icon.py {data_dir} -d OfficeHome -s Ar -t Cl -a resnet50 --epochs 50 --lr 0.005 --temperature 2.5 --bottleneck-dim 2048  --w-transfer 0.015 --w-st 0.5 --threshold 0.97 --log-root {log_dir} --batch-size 28  --con-start-epoch 0 --con-mode stats --back-cluster-start-epoch 0 --topk 5 --seed 0 --w-inv 0.1 --inv-start-epoch 10 --exp-name Ar2Cl --optim sam

Acknowledgement

This code is implemented based on the CST, and it is our pleasure to acknowledge their contributions.

Citation

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

@article{yue2023make,
  title={Make the U in UDA Matter: Invariant Consistency Learning for Unsupervised Domain Adaptation},
  author={Yue, Zhongqi and Sun, Qianru and Zhang, Hanwang},
  journal={Advances in neural information processing systems},
  year={2023}
}

Contact

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

• [email protected]