TD3Net: A temporal densely connected multi-dilated convolutional network for lipreading

Byung Hoon Lee, Wooseok Shin, Sung Won Han

The word-level lipreading approach typically employs a two-stage framework with separate frontend and backend architectures to model dynamic lip movements. Each component has been…