Bi-discriminator Domain Adversarial Neural Networks with Class-Level Gradient Alignment

About Our Work

Update: 2023/11/23: We have created a repository for the paper titled Bi-discriminator Domain Adversarial Neural Networks with Class-Level Gradient Alignment, which has been submitted to the **IEEE Transactions on Systems, Man, and Cybernetics: Systems (SMCA) **. In this repository, we offer the original sample datasets, preprocessing scripts, and algorithm files to showcase the reproducibility of our work.

Requirements

• Python == 3.8.10
• Pytorch == 1.2.0
• timm== 0.9.11
• scikit-learn== 1.2.2
• wilds== 2.0.0

Data Sets

The structure of the data set should be like

data
|_ clef
|  |_ b
|  |_ c
|  |_ i
|  |_ p
|  |_ list
|_ digit
|_ |_ ...
|_ visda
|_ |_ ...
|_ cifa
|_ |_ ...

Due to the copyright limitations, we have not uploaded the data. You can seek permission from the organizer according to the link given or download it directly from their website.

ImageClef Dataset

CIFAR Dataset

Visda2017 Dataset

Digits Dataset

RUN

You should update the log and data reading directories in the configuration file initially.

# unzip all files into the DA directory
# run BACG
python main.py
# run Fast-BACG
python main_mem.py

Contributors ✨

Many thanks to the data preprocessing pipeline in the following published papers.

Transfer Learning Library； Deep Evidential Learning; CDGM