Skip to content

Pengchengpcx/Class-overwhelms-Mutual-Conditional-Blended-Target-Domain-Adaptation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Class-Overwhelms-Mutual-Conditional-Blended-Target-Domain-Adaptation

Official Implementation for the AAAI-2023 Oral paper

Pengcheng Xu, Boyu Wang, Charles Ling, Class-Overwhelms-Mutual-Conditional-Blended-Target-Domain-Adaptation

Prerequisites:

  • python == 3.9.6
  • pytorch ==1.12.1
  • torchvision == 0.13.1
  • numpy, scipy, sklearn, PIL, argparse, tqdm
  • RandAugment

Data Preparation

  • Please manually download the datasets Office, Office-Home, DomainNet

  • Add the system path into preparedata_lds.py and preparedata_uda.py in datasets and cada_styflip.py in train.

  • Create the log directory for each dataset

./logs/office-home-btlds
./logs/domainnet
./logs/office-home

Training

You can run the training file in train with

python train/cada_styflip.py --dataset office-home-btlds --bs_limit 64 --iter_epoch 500 --source 0 --catal --batch_size 1 --sub_log styflip_btlds_noaug --amp

python train/cada_styflip.py --feat_dim 1024 --hid_dim 2048 --dataset domainnet --net resnet101 --iter_epoch 800 --source 0 --catal --batch_size 1 --bs_limit 256 --max_epoch 10 --sub_log styflip --amp


Framework

If you find our work useful in your research, please consider citing:

@article{xu2023class,
  title={Class Overwhelms: Mutual Conditional Blended-Target Domain Adaptation},
  author={Xu, Pengcheng and Wang, Boyu and Ling, Charles},
  journal={arXiv preprint arXiv:2302.01516},
  year={2023}
}

License

This repository is released under MIT License (see LICENSE file for details).

About

[AAAI2023] Class Overwhelms: Mutual Conditional Blended-Target Domain Adaptation

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages