/MAE-CIL

Official implementation of the paper "Masked Autoencoders are Efficient Class Incremental Learners"

Primary LanguagePython

MAE-CIL - Official PyTorch Implementation

[ICCV 2023] Masked Autoencoders are Efficient Class Incremental Learners

Paper

Usage

  • Training on CIFAR-100 dataset:
$ sh sv.sh

Environment

  • We recommend you to use python3.6 as the base python environment. And other dependency libraries are listed in requirements.txt.
$ pip install -r requirements.txt

Data Download

  • Please download the MAE buffer used for training. After download it, please put it in the root folder.

Acknowledge

  • We use the codebase from DyTox to align some data processing implementations and evaluation metrics. Thanks for their wonderful work!

Citation

If you use this code for your research, please consider citing:

@inproceedings{zhai2023masked,
  title={Masked autoencoders are efficient class incremental learners},
  author={Zhai, Jiang-Tian and Liu, Xialei and Bagdanov, Andrew D and Li, Ke and Cheng, Ming-Ming},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={19104--19113},
  year={2023}
}