Research map / Vision-language and multimodal / Cross-modal learning
Deep Learning for Visual Speech Analysis: A Survey
Changchong Sheng, Gangyao Kuang, Liang Bai, Chenping Hou, Yulan Guo, Xin Xu, Matti Pietikäinen, Li Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use multi-stage knowledge distillation | C3D-ResNet18 + MS-TCN · LRW | accuracy: 85.3 -> 87.7 | Improved |
| add cross-modal mutual learning with audio in training | C3D-ResNet18 · LRW | accuracy: 87.7 -> 88.5 | Improved |
| add cross-modal mutual learning with audio in training | C3D-ResNet18 · LRW-1000 | accuracy: 48.0 -> 50.5 | Improved |
| use seq2seq loss instead of CTC loss | Afouras et al. Transformer · LRS2 | CER: 54.7 -> 48.3 | Improved |
| use seq2seq loss instead of CTC loss | Afouras et al. Transformer · LRS3 | WER: 66.3 -> 58.9 | Improved |
| replace 2D CNN with visual transformer and sub-word output | Prajwal et al. · LRS3 | WER: 43.3 -> 30.7 | Improved |
| replace 2D CNN with visual transformer and sub-word output | Prajwal et al. · LRS2 | WER: 37.9 -> 22.6 | Improved |
| use BiGRU backend with C3D front end instead of ST-CNN CTC | Xu et al. · GRID | WER: 4.8 -> 2.9 | Improved |
| use MS-TCN backend instead of BiLSTM | Martinez et al. · LRW | accuracy: 83.5 -> 85.3 | Improved |
Closest papers by meaning
- Visual Hallucination Elevates Speech Recognition (AAAI 2024 · similarity 0.85)
- VSRo-200: A Romanian Visual Speech Recognition Dataset for Studying Supervision and Multimodal Robustness (NeurIPS 2026 · similarity 0.85)
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · similarity 0.85)
- SVHalluc: Benchmarking Speech-Vision Hallucination in Audio-Visual Large Language Models (CVPR 2026 · similarity 0.84)
- video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models (ICML 2024 · similarity 0.84)
- Vision-Speech Models: Teaching Speech Models to Converse about Images (CVPR 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
