Estimating (and Fixing) the Effect of Face Obfuscation in Video Recognition

Matteo Tomei, Lorenzo Baraldi, Simone Bronzin, Rita Cucchiara

CVPRW 2021 pp. 3263-3269

doi:10.1109/CVPRW53098.2021.00364 /cvprw/2021/tomei2021cvprw-estimating/

Abstract

Recent research has shown that faces can be obfuscated in large-scale datasets with a minimal performance impact on image classification and downstream tasks like object recognition. In this paper, we investigate the role of face obfuscation in video classification datasets and quantify a more significant reduction in performance caused by face blurring. To reduce such performance effects, we propose a generalized distillation approach in which a privacy-preserving action recognition network is trained with privileged information given by face identities. We show, through experiments performed on Kinetics-400, that the proposed approach can fully close the performance gap caused by face anonymization.

CVPRW Semantic Scholar

Cite

Text

Tomei et al. "Estimating (and Fixing) the Effect of Face Obfuscation in Video Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021. doi:10.1109/CVPRW53098.2021.00364

Markdown

[Tomei et al. "Estimating (and Fixing) the Effect of Face Obfuscation in Video Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2021.](https://mlanthology.org/cvprw/2021/tomei2021cvprw-estimating/) doi:10.1109/CVPRW53098.2021.00364

BibTeX

@inproceedings{tomei2021cvprw-estimating,
  title     = {{Estimating (and Fixing) the Effect of Face Obfuscation in Video Recognition}},
  author    = {Tomei, Matteo and Baraldi, Lorenzo and Bronzin, Simone and Cucchiara, Rita},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2021},
  pages     = {3263-3269},
  doi       = {10.1109/CVPRW53098.2021.00364},
  url       = {https://mlanthology.org/cvprw/2021/tomei2021cvprw-estimating/}
}