Angular Gap

This is the code necessary to run experiments described in the ACM MM'22 paper Paper

Requirements

All the required packages can be installed by running pip install -r requirements.txt.

Or pull our domain adaptation docker image and run experiments with

docker pull marvinpeng2022/da-testbed

Difficulty estimation

python main.py --dst cifar10 --arch resnet18

Visualization

python main.py --dst cifar10 --arch visualization

Domain adaptation

For domain adaptation, we have released our implementation of CRST and Curricular DSAN.

cd DeepDA
bash run.sh

Video

Presentation and Slides

If you make use of this code in your work, please cite the following paper:

@article{peng2022angular,
  title={Angular Gap: Reducing the Uncertainty of Image Difficulty through Model Calibration},
  author={Peng, Bohua and Islam, Mobarakol and Tu, Mei},
  journal={arXiv preprint arXiv:2207.08525},
  year={2022}
}