Audio Barlow Twins: Self-Supervised Audio Representation Learning

The Barlow Twins self-supervised learning objective requires neither negativesamples or asymmetric learning updates, achieving results on a par with thecurrent state-of-the-art within Computer Vision. As such, we present AudioBarlow Twins, a novel self-supervised audio representation learning approach,adapting Barlow Twins to the audio domain. We pre-train on the large-scaleaudio dataset AudioSet, and evaluate the quality of the learnt representationson 18 tasks from the HEAR 2021 Challenge, achieving results which outperform,or otherwise are on a par with, the current state-of-the-art for instancediscrimination self-supervised learning approaches to audio representationlearning. Code at https://github.com/jonahanton/SSL_audio.