Temporal Action Localization in Untrimmed Videos via Multi-stage CNNs

TL;DR

This paper presents a multi-stage CNN framework for accurately localizing actions in untrimmed videos by proposing candidate segments, classifying actions, and fine-tuning localization with a novel loss function, achieving state-of-the-art results.

Contribution

It introduces a three-stage CNN approach with a new loss function for improved temporal action localization in untrimmed videos.

Findings

Significant performance improvements on MEXaction2 and THUMOS 2014 datasets.

Proposed method increases mAP from 1.7% to 7.4% on MEXaction2.

Achieves 19.0% mAP on THUMOS 2014 at 0.5 overlap threshold.

Abstract

We address temporal action localization in untrimmed long videos. This is important because videos in real applications are usually unconstrained and contain multiple action instances plus video content of background scenes or other activities. To address this challenging issue, we exploit the effectiveness of deep networks in temporal action localization via three segment-based 3D ConvNets: (1) a proposal network identifies candidate segments in a long video that may contain actions; (2) a classification network learns one-vs-all action classification model to serve as initialization for the localization network; and (3) a localization network fine-tunes on the learned classification network to localize each action instance. We propose a novel loss function for the localization network to explicitly consider temporal overlap and therefore achieve high temporal localization accuracy. Only the proposal network and the localization network are used during prediction. On two large-scale benchmarks, our approach achieves significantly superior performances compared with other state-of-the-art systems: mAP increases from 1.7% to 7.4% on MEXaction2 and increases from 15.0% to 19.0% on THUMOS 2014, when the overlap threshold for evaluation is set to 0.5.