Fetching the paper…

Learning Spatio-Temporal Features with Two-Stream Deep 3D CNNs for Lipreading · Around