/Code_ICPR

Primary LanguagePython

Reproducible Code for our work Intrinsic Consistency Preservation with Adaptively Reliable Sample for Source-free Domain Adaptation.

Installation and requirements

python == 3.7
pytorch == 1.10
cudatoolkit == 11.1
torchvision == 0.11

Data preparation

Office, Office-Home, VisDA-C, Office-Home (RSUT), VisDA-C (RSUT), and DomainNet can be found in the related repo ISFDA. Note that these datasets are publicly available and have been widely used in the related research community.

Training and adaptation

Taking the VisDA dataset as an example, the training command is:

bash image_source.sh

The adaptation command is:

bash image_target.sh

Acknowledge

Our code partially follows the programming style of NRC and SHOT. Thanks for the excellent work of NRC and SHOT.