Delta Sampling R-BERT for Limited Data and Low-Light Action Recognition

Sanchit Hira, Ritwik Das, Abhinav Modi, Daniil Pakhomov

CVPRW 2021 pp. 853-862

doi:10.1109/CVPRW53098.2021.00095 /cvprw/2021/hira2021cvprw-delta/

Abstract

We present an approach to perform supervised action recognition in the dark. In this work, we present our results on the ARID dataset [60]. Most previous works only evaluate performance on large, well illuminated datasets like Kinetics and HMDB51. We demonstrate that our work is able to achieve a very low error rate while being trained on a much smaller dataset of dark videos. We also explore a variety of training and inference strategies including domain transfer methodologies and also propose a simple but useful frame selection strategy. Our empirical results demonstrate that we beat previously published baseline models by 11%.

PDF CVPRW Semantic Scholar

Cite

Text

Hira et al. "Delta Sampling R-BERT for Limited Data and Low-Light Action Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021. doi:10.1109/CVPRW53098.2021.00095

Markdown

[Hira et al. "Delta Sampling R-BERT for Limited Data and Low-Light Action Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021.](https://mlanthology.org/cvprw/2021/hira2021cvprw-delta/) doi:10.1109/CVPRW53098.2021.00095

BibTeX

@inproceedings{hira2021cvprw-delta,
  title     = {{Delta Sampling R-BERT for Limited Data and Low-Light Action Recognition}},
  author    = {Hira, Sanchit and Das, Ritwik and Modi, Abhinav and Pakhomov, Daniil},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2021},
  pages     = {853-862},
  doi       = {10.1109/CVPRW53098.2021.00095},
  url       = {https://mlanthology.org/cvprw/2021/hira2021cvprw-delta/}
}