EPIC-Fusion: Audio-Visual Temporal Binding for Egocentric Action Recognition

TL;DR

EPIC-Fusion introduces a novel multi-modal architecture for egocentric action recognition that effectively combines RGB, Flow, and Audio modalities through temporal-binding and mid-level fusion, achieving state-of-the-art results.

Contribution

The paper proposes a new end-to-end multi-modal fusion architecture that fuses modalities before temporal aggregation, enhancing egocentric action recognition performance.

Findings

Audio significantly improves action and object interaction recognition.

The method outperforms previous approaches on EPIC-Kitchens dataset.

Achieves state-of-the-art results on all metrics for seen and unseen test sets.

Abstract

We focus on multi-modal fusion for egocentric action recognition, and propose a novel architecture for multi-modal temporal-binding, i.e. the combination of modalities within a range of temporal offsets. We train the architecture with three modalities -- RGB, Flow and Audio -- and combine them with mid-level fusion alongside sparse temporal sampling of fused representations. In contrast with previous works, modalities are fused before temporal aggregation, with shared modality and fusion weights over time. Our proposed architecture is trained end-to-end, outperforming individual modalities as well as late-fusion of modalities. We demonstrate the importance of audio in egocentric vision, on per-class basis, for identifying actions as well as interacting objects. Our method achieves state of the art results on both the seen and unseen test sets of the largest egocentric dataset: EPIC-Kitchens, on all metrics using the public leaderboard.