Speed Co-Augmentation for Unsupervised Audio-Visual Pre-training

TL;DR

This paper introduces a simple yet effective speed co-augmentation technique for unsupervised audio-visual pre-training, which enhances data diversity and representation quality by altering playback speeds and employing a novel loss function.

Contribution

It proposes a novel speed co-augmentation method combined with SoftInfoNCE loss to improve unsupervised audio-visual representation learning.

Findings

Significant improvement over vanilla contrastive learning

Increases diversity and negative pairs in training

Enhances learned representations with simple augmentation

Abstract

This work aims to improve unsupervised audio-visual pre-training. Inspired by the efficacy of data augmentation in visual contrastive learning, we propose a novel speed co-augmentation method that randomly changes the playback speeds of both audio and video data. Despite its simplicity, the speed co-augmentation method possesses two compelling attributes: (1) it increases the diversity of audio-visual pairs and doubles the size of negative pairs, resulting in a significant enhancement in the learned representations, and (2) it changes the strict correlation between audio-visual pairs but introduces a partial relationship between the augmented pairs, which is modeled by our proposed SoftInfoNCE loss to further boost the performance. Experimental results show that the proposed method significantly improves the learned representations when compared to vanilla audio-visual contrastive learning.