Conformers are All You Need for Visual Speech Recognition

Benchmark Model Rank Results
lipreading-on-lrs3-tedLP + Conformer–Word Error Rate (WER): 12.8