A Multimodal German Dataset for Automatic Lip Reading Systems and Transfer Learning

TL;DR

This paper introduces the GLips dataset, a large German lip reading dataset, and explores transfer learning between German and English lip reading datasets to improve model performance.

Contribution

The paper presents the new GLips dataset and demonstrates the effectiveness of transfer learning between German and English lip reading datasets.

Findings

Transfer learning improves lip reading performance across languages.

GLips enables studying language-independent features in lip reading.

Models trained with transfer learning show faster learning and higher accuracy.

Abstract

Large datasets as required for deep learning of lip reading do not exist in many languages. In this paper we present the dataset GLips (German Lips) consisting of 250,000 publicly available videos of the faces of speakers of the Hessian Parliament, which was processed for word-level lip reading using an automatic pipeline. The format is similar to that of the English language LRW (Lip Reading in the Wild) dataset, with each video encoding one word of interest in a context of 1.16 seconds duration, which yields compatibility for studying transfer learning between both datasets. By training a deep neural network, we investigate whether lip reading has language-independent features, so that datasets of different languages can be used to improve lip reading models. We demonstrate learning from scratch and show that transfer learning from LRW to GLips and vice versa improves learning speed and performance, in particular for the validation set.