RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization

RefineLoc is a weakly-supervised action detector that aims to approximate the true foreground-background labels through iteratively generating pseudo ground truth. Our key idea is to use the pseudo ground truth from iteration n-1 to supervise the detection model at iteration n


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 labelling, where videos are untrimmed and only a video-level label is available. In this paper, we propose RefineLoc, a new weakly-supervised 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 different pseudo ground truth generators. We show the effectiveness of our model on two standard action datasets, ActivityNet v1.2 and THUMOS14. RefineLoc equipped with a segment prediction-based pseudo ground truth generator improves the state-of-the-art in weakly-supervised temporal localization on the challenging and large-scale ActivityNet dataset by 1.5% in average mAP.

On arXiv


    title={RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization},
    author={Humam Alwassel and Alejandro Pardo and Fabian Caba Heilbron and Ali Thabet and Bernard Ghanem},