QinghaoLi0421/DCA_code

Code for our article Dual Classifier Adaptation: Source-Free UDA via Adaptive Pseudo-labels Learning

DCA

Code in Pytorch for our article Dual Classifier Adaptation: Source-Free UDA via Adaptive Pseudo-labels Learning

Environment Requirement

• python 3.8.10
• pytorch 1.11.0
• torchvision 0.12.0
• cuda 10.1
• numpy, sklearn, scipy, tqdm

Dataset

• Please manually download the datasets Office, Office-Home, VisDA-Cfrom the official websites, and modify the path of images in each '.txt' under the folder './data/'. [**How to generate such txt files could be found in https://github.com/tim-learn/Generate_list **]

Training

Firstly, you need to train the source model by running this code

python train_source_31.py

Then, you can run the code of DCA to perform SFDA,

python train_target_31.py