DNE

This is the official implementation for "Dense Network Expansion for Class Incremental Learning" (CVPR2023)

Environment setup

The dependencies are included in requirements.txt, which can be installed through:

pip install -r requirements.txt

Dataset preparation

Please use the --data-path parameter in the bash file to config the root directory of your datasets.

CIFAR100 dataset will be automatically downloaded if it does not exist in the dataset root.

ImageNet needs be manually downloaded to the dataset root.

Training and evaluation

Please use train.sh and train_imagenet100.sh to train and evaluate the model on CIFAR100 and ImageNet100 datasets.

Acknowledgement

This repo owes a huge thanks to the dytox repo.

Citation

If this work is helpful to you, please cite our work as:

@inproceedings{hu2023dense,
  title={Dense network expansion for class incremental learning},
  author={Hu, Zhiyuan and Li, Yunsheng and Lyu, Jiancheng and Gao, Dashan and Vasconcelos, Nuno},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={11858--11867},
  year={2023}
}

Contact

Feel free to contact [email protected] if you have any question about this repo or our work.

Future updates

We will continue polish this code repo to make it easier to read and reproduce.