LipNet: End-to-End Sentence-level Lipreading

TL;DR

LipNet is an innovative end-to-end deep learning model that accurately predicts entire sentences from lip movements, outperforming previous models and human lipreaders on a standard dataset.

Contribution

LipNet is the first end-to-end model for sentence-level lipreading that jointly learns visual features and sequence modeling using spatiotemporal convolutions and CTC loss.

Findings

Achieves 95.2% accuracy on GRID corpus sentence task

Outperforms previous word-level state-of-the-art and human lipreaders

Demonstrates the effectiveness of end-to-end training for lipreading

Abstract

Lipreading is the task of decoding text from the movement of a speaker's mouth. Traditional approaches separated the problem into two stages: designing or learning visual features, and prediction. More recent deep lipreading approaches are end-to-end trainable (Wand et al., 2016; Chung & Zisserman, 2016a). However, existing work on models trained end-to-end perform only word classification, rather than sentence-level sequence prediction. Studies have shown that human lipreading performance increases for longer words (Easton & Basala, 1982), indicating the importance of features capturing temporal context in an ambiguous communication channel. Motivated by this observation, we present LipNet, a model that maps a variable-length sequence of video frames to text, making use of spatiotemporal convolutions, a recurrent network, and the connectionist temporal classification loss, trained entirely end-to-end. To the best of our knowledge, LipNet is the first end-to-end sentence-level lipreading model that simultaneously learns spatiotemporal visual features and a sequence model. On the GRID corpus, LipNet achieves 95.2% accuracy in sentence-level, overlapped speaker split task, outperforming experienced human lipreaders and the previous 86.4% word-level state-of-the-art accuracy (Gergen et al., 2016).