RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization

Abstract
Video action detectors are usually trained using datasets with fully-supervised temporal annotations. Building such datasets is an expensive task. To alleviate this problem, recent methods have tried to leverage weak labeling, where videos are untrimmed and only a video-level label is available. In this paper, we propose RefineLoc, a novel weaklysupervised temporal action localization method. RefineLoc uses an iterative refinement approach by estimating and training on snippet-level pseudo ground truth at every iteration. We show the benefit of this iterative approach and present an extensive analysis of five different pseudo ground truth generators. We show the effectiveness of our model on two standard action datasets, ActivityNet v1.2 and THUMOS14. RefineLoc shows competitive results with the stateof-the-art in weakly-supervised temporal localization. Additionally, our iterative refinement process is able to significantly improve the performance of two state-of-the-art methods, setting a new state-of-the-art on THUMOS14.

Citation
Pardo, A., Alwassel, H., Heilbron, F. C., Thabet, A., & Ghanem, B. (2021). RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization. 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). doi:10.1109/wacv48630.2021.00336

Acknowledgements
This work is supported the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No. OSRCRG2017-3405.

Publisher
IEEE

Conference/Event Name
2021 IEEE Winter Conference on Applications of Computer Vision (WACV)

DOI
10.1109/WACV48630.2021.00336

arXiv
1904.00227

Additional Links
https://ieeexplore.ieee.org/document/9423165/
https://openaccess.thecvf.com/content/WACV2021/papers/Pardo_RefineLoc_Iterative_Refinement_for_Weakly-Supervised_Action_Localization_WACV_2021_paper.pdf

Relations
Is Supplemented By:
  • [Software]
    Title: HumamAlwassel/RefineLoc: RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization (WACV 2021). Publication Date: 2021-01-03. github: HumamAlwassel/RefineLoc Handle: 10754/669741

Permanent link to this record

Version History

Now showing 1 - 2 of 2
VersionDateSummary
2*
2021-06-16 06:37:09
Published as conference paper.
2019-12-18 11:23:14
* Selected version