SpotFast Networks with Memory Augmented Lateral Transformers for Lipreading
This work addresses lipreading for applications like assistive technology, but it is incremental as it builds on existing SlowFast networks.
The paper tackles word-level lipreading by introducing SpotFast networks, a variant of SlowFast networks, combined with memory augmented lateral transformers, achieving a 3.7% improvement over the base model on the LRW dataset.
This paper presents a novel deep learning architecture for word-level lipreading. Previous works suggest a potential for incorporating a pretrained deep 3D Convolutional Neural Networks as a front-end feature extractor. We introduce a SpotFast networks, a variant of the state-of-the-art SlowFast networks for action recognition, which utilizes a temporal window as a spot pathway and all frames as a fast pathway. We further incorporate memory augmented lateral transformers to learn sequential features for classification. We evaluate the proposed model on the LRW dataset. The experiments show that our proposed model outperforms various state-of-the-art models and incorporating the memory augmented lateral transformers makes a 3.7% improvement to the SpotFast networks.