Hear Me Out: Fusional Approaches for Audio Augmented Temporal Action Localization

TL;DR

This paper introduces fusion-based methods that incorporate audio information into temporal action localization in videos, significantly improving performance on large-scale benchmarks and pioneering the joint use of audio and video modalities for this task.

Contribution

It is the first to jointly consider audio and video modalities for supervised TAL, demonstrating consistent performance improvements over video-only approaches.

Findings

Achieved new state-of-the-art results on ActivityNet-1.3 and THUMOS14 datasets.

Multiple fusion schemes and modality combinations improve TAL accuracy.

Experimental ablations validate the effectiveness of audio-video fusion in TAL.

Abstract

State of the art architectures for untrimmed video Temporal Action Localization (TAL) have only considered RGB and Flow modalities, leaving the information-rich audio modality totally unexploited. Audio fusion has been explored for the related but arguably easier problem of trimmed (clip-level) action recognition. However, TAL poses a unique set of challenges. In this paper, we propose simple but effective fusion-based approaches for TAL. To the best of our knowledge, our work is the first to jointly consider audio and video modalities for supervised TAL. We experimentally show that our schemes consistently improve performance for state of the art video-only TAL approaches. Specifically, they help achieve new state of the art performance on large-scale benchmark datasets - ActivityNet-1.3 (54.34 mAP@0.5) and THUMOS14 (57.18 mAP@0.5). Our experiments include ablations involving multiple fusion schemes, modality combinations and TAL architectures. Our code, models and associated data are available at https://github.com/skelemoa/tal-hmo.