Awesome

Wakening Past Concepts without Past Data: <br>Class-Incremental Learning from Online Placebos

This repository contains the PyTorch implementation for the WACV 2024 Paper "Wakening Past Concepts without Past Data: Class-Incremental Learning from Online Placebos".

Initialize the environment

conda create --name onlineplacebos python=3.8
conda activate onlineplacebos
conda install cudatoolkit=11.1 -c nvidia
conda install pytorch torchvision torchaudio -c pytorch-lts

pip install tqdm 
pip install scipy 
pip install quadprog
pip install POT

Run the code

bash run_exp_all.py

Citation

Please cite our paper if it is helpful to your work:

@inproceedings{Liu2024OnlinePlacebos,
  author       = {Yaoyao Liu and
                  Yingying Li and
                  Bernt Schiele and
                  Qianru Sun},
  title        = {Wakening Past Concepts without Past Data: Class-Incremental Learning
                  from Online Placebos},
  booktitle    = {{IEEE/CVF} Winter Conference on Applications of Computer Vision, {WACV}
                  2024, Waikoloa, HI, USA, January 3-8, 2024},
  pages        = {2215--2224},
  publisher    = {{IEEE}},
  year         = {2024},
  url          = {https://doi.org/10.1109/WACV57701.2024.00222}
}

Acknowledgements

Our implementation uses the source code from the following repository:

• https://github.com/G-U-N/PyCIL