Visual Speech Recognition for Multiple Languages in the Wild

Benchmark Model Rank Results
lipreading-on-cmlrCTC/Attention#1CER: 9.1%
lipreading-on-grid-corpus-mixed-speechCTC/Attention#1Word Error Rate (WER): 1.2
lipreading-on-lrs2CTC/Attention (LRW+LRS2/3+AVSpeech)#5Word Error Rate (WER): 25.5
lipreading-on-lrs2CTC/Attention#7Word Error Rate (WER): 32.9
lipreading-on-lrs3-tedCTC/Attention (LRW+LRS2/3+AVSpeech)#11Word Error Rate (WER): 31.5