This paper proposes a powerful Visual Speech Recognition (VSR) method for multiple languages, especially for low-resource languages that have a limited number of labeled data. Different from previous ...

We used 3 publicly available datasets in this work. The Lip Reading in the Wild (LRW) dataset was used for visual pre-training and the Lip Reading Sentences 2 (LRS2) and Lip Reading Sentences 3 (LRS3) ...