Audio-Visual Synchronisation in the wild

TL;DR

This paper introduces a new dataset and benchmark for general audio-visual synchronisation across diverse video classes, proposing transformer-based models that outperform previous methods.

Contribution

The paper presents the first benchmark for open domain audio-visual synchronisation and develops transformer models tailored for arbitrary-length signals with reduced memory use.

Findings

Proposed models outperform previous state-of-the-art methods.

Curated a new dataset VGG-Sound Sync with high audio-visual correlation.

Set a new benchmark for diverse audio-visual synchronisation tasks.

Abstract

In this paper, we consider the problem of audio-visual synchronisation applied to videos `in-the-wild' (ie of general classes beyond speech). As a new task, we identify and curate a test set with high audio-visual correlation, namely VGG-Sound Sync. We compare a number of transformer-based architectural variants specifically designed to model audio and visual signals of arbitrary length, while significantly reducing memory requirements during training. We further conduct an in-depth analysis on the curated dataset and define an evaluation metric for open domain audio-visual synchronisation. We apply our method on standard lip reading speech benchmarks, LRS2 and LRS3, with ablations on various aspects. Finally, we set the first benchmark for general audio-visual synchronisation with over 160 diverse classes in the new VGG-Sound Sync video dataset. In all cases, our proposed model outperforms the previous state-of-the-art by a significant margin.