LiRA: Learning Visual Speech Representations from Audio through Self-supervision

TL;DR

LiRA introduces a self-supervised learning approach that trains visual speech representations from audio to improve lip-reading accuracy, outperforming existing methods on standard datasets with less labeled data.

Contribution

The paper presents a novel self-supervised method to learn visual speech features from audio, enhancing lip-reading performance with limited labeled data.

Findings

Outperforms other self-supervised methods on LRW dataset

Achieves state-of-the-art on LRS2 with less labeled data

Effective transfer from audio-based pretraining to visual lip-reading

Abstract

The large amount of audiovisual content being shared online today has drawn substantial attention to the prospect of audiovisual self-supervised learning. Recent works have focused on each of these modalities separately, while others have attempted to model both simultaneously in a cross-modal fashion. However, comparatively little attention has been given to leveraging one modality as a training objective to learn from the other. In this work, we propose Learning visual speech Representations from Audio via self-supervision (LiRA). Specifically, we train a ResNet+Conformer model to predict acoustic features from unlabelled visual speech. We find that this pre-trained model can be leveraged towards word-level and sentence-level lip-reading through feature extraction and fine-tuning experiments. We show that our approach significantly outperforms other self-supervised methods on the Lip Reading in the Wild (LRW) dataset and achieves state-of-the-art performance on Lip Reading Sentences 2 (LRS2) using only a fraction of the total labelled data.