VGGSound: A Large-scale Audio-Visual Dataset

TL;DR

VGGSound is a large-scale, low-noise audio-visual dataset created using a scalable computer vision pipeline, enabling improved audio recognition research with over 210,000 videos across 310 classes.

Contribution

The paper introduces a novel scalable pipeline for collecting large-scale audio-visual data from videos in the wild, resulting in the VGGSound dataset with high-quality audio-visual correspondence.

Findings

Established baseline audio recognition models on VGGSound

Demonstrated the dataset's suitability for training deep learning models

Compared VGGSound with existing datasets to highlight its advantages

Abstract

Our goal is to collect a large-scale audio-visual dataset with low label noise from videos in the wild using computer vision techniques. The resulting dataset can be used for training and evaluating audio recognition models. We make three contributions. First, we propose a scalable pipeline based on computer vision techniques to create an audio dataset from open-source media. Our pipeline involves obtaining videos from YouTube; using image classification algorithms to localize audio-visual correspondence; and filtering out ambient noise using audio verification. Second, we use this pipeline to curate the VGGSound dataset consisting of more than 210k videos for 310 audio classes. Third, we investigate various Convolutional Neural Network~(CNN) architectures and aggregation approaches to establish audio recognition baselines for our new dataset. Compared to existing audio datasets, VGGSound ensures audio-visual correspondence and is collected under unconstrained conditions. Code and the dataset are available at http://www.robots.ox.ac.uk/~vgg/data/vggsound/