TIM: A Time Interval Machine for Audio-Visual Action Recognition

TL;DR

The paper introduces TIM, a novel transformer-based model that explicitly models the temporal extents of audio and visual events in long videos, achieving state-of-the-art results in action recognition and detection.

Contribution

TIM is the first model to explicitly incorporate modality-specific time intervals for audio-visual action recognition in long videos, improving accuracy and detection performance.

Findings

TIM outperforms previous SOTA on EPIC-KITCHENS with 2.9% higher top-1 accuracy.

TIM achieves superior results in action detection using dense multi-scale interval queries.

Integrating both modalities and modeling their time intervals is crucial for high performance.

Abstract

Diverse actions give rise to rich audio-visual signals in long videos. Recent works showcase that the two modalities of audio and video exhibit different temporal extents of events and distinct labels. We address the interplay between the two modalities in long videos by explicitly modelling the temporal extents of audio and visual events. We propose the Time Interval Machine (TIM) where a modality-specific time interval poses as a query to a transformer encoder that ingests a long video input. The encoder then attends to the specified interval, as well as the surrounding context in both modalities, in order to recognise the ongoing action. We test TIM on three long audio-visual video datasets: EPIC-KITCHENS, Perception Test, and AVE, reporting state-of-the-art (SOTA) for recognition. On EPIC-KITCHENS, we beat previous SOTA that utilises LLMs and significantly larger pre-training by 2.9% top-1 action recognition accuracy. Additionally, we show that TIM can be adapted for action detection, using dense multi-scale interval queries, outperforming SOTA on EPIC-KITCHENS-100 for most metrics, and showing strong performance on the Perception Test. Our ablations show the critical role of integrating the two modalities and modelling their time intervals in achieving this performance. Code and models at: https://github.com/JacobChalk/TIM