Growing a Brain with Sparsity-Inducing Generation for Continual Learning

This repos contains code for continually training video action recognition task from our Growing a Brain with Sparsity-Inducing Generation for Continual Learning (ICCV 2023). Please see our paper for more detailed information.

Requirements

Before running the code, please install the requirements listed in the requirements.txt file.

Run the code

This repository supports the video action recognition experiment with UCF-101 in the original paper.

python3 -u ucf_main.py | tee growbrain.log

Before running the codes, you have to download the video datasets and extract the frames of videos. We followed the video action recognition benchmark provided from [vCLIMB]. Each video is split into three segments of equal duration. In each segment, a frame is selected randomly.

Citation

@inproceedings{jin2023growing,
  title={Growing a Brain with Sparsity-Inducing Generation for Continual Learning},
  author={Jin, Hyundong and Kim, Gyeong-hyeon and Ahn, Chanho and Kim, Eunwoo},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={18961--18970},
  year={2023}
}