/TAF-Cal

Test-time Fourier Style Calibration for Domain Generalization - IJCAI 2022

Primary LanguagePython

Test-time Fourier Style Calibration for Domain Generalization

Paper | IJCAI-ECAI 2022 | SMiLe Lab

Xingchen Zhao1, Chang Liu1, Anthony Sicilia2, Seong Jae Hwang2, Yun Raymond Fu1

1 Northeastern University 2 University of Pittsburgh

This is the official implementation of the paper "Test-time Fourier Style Calibration for Domain Generalization". The code will cleaned up soon. Please contact zhao.xingc@northeastern.edu or create an issue if you have any questions. Thank you for your interest.

Install Conda Environment

Please visit the Anaconda install page if you do not already have conda installed

conda env create --name TAF_Cal --file=environment.yml
conda activate TAF_Cal

Download Dataset

bash download_dataset.sh

Training and Evaluation

The evaluation stage is included in the training code.

To train the deepall, please run:

bash pacs_deepall.sh

To train TAF_Cal, please run (Comparing the results of the paper, the result produced by this code will fluctuate 0.5~0.8% up or down due to the random seed. (Try multiple random seeds and take the average):

bash pacs_taf_cal_train.sh

Reference

If our work or code helps you, please consider to cite our paper. Thank you!

@inproceedings{wang2022r2l,
  author = {Xingchen Zhao and Chang Liu and Anthony Scilia and Seong Jae Hwang and Yun Fu},
  title = {Test-time Fourier Style Calibration for Domain Generalization},
  booktitle = {IJCAI},
  year = {2022}
}