Research map / Video / Video segmentation and tracking
A First Exploration of Neuromorphic OT-CFM for Multi-Speaker VSR
Lin Chen, Jingping Fang, Hairui Liu, Chenyang Xu, Junhao Chen, Xiaorui Li, Weidong Cai, Xiaoming Chen
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 |
|---|---|---|---|
| add event-stream branch via cross-attention fusion to visual-only (ablation) | LipsFlow (AR decoder) · DVS-Lip | WER: 27.8 -> 24.5 | Improved |
| use cross-attention instead of naive channel concat fusion (ablation) | LipsFlow (AR decoder) · DVS-Lip | WER: 25.9 -> 24.5 | Improved |
| replace AR decoder with OT-CFM decoder (ablation) | LipsFlow · DVS-Lip | WER: 24.5 -> 22.3 | Improved |
| replace diffusion decoder with OT-CFM (ablation) | LipsFlow · DVS-Lip | WER: 22.5 -> 22.3 | No clear effect |
| increase ODE steps from 1 to 2 NFE (ablation) | LipsFlow OT-CFM · DVS-Lip | WER: 28.5 -> 22.3 | Improved |
| increase ODE steps from 2 to 10 NFE (ablation) | LipsFlow OT-CFM · DVS-Lip | WER: 22.3 -> 22.0 | No clear effect |
| replace CE-only objective with OT-CFM (ablation) | LipsFlow · DVS-Lip | WER: 35.8 -> 28.2 | Improved |
| add speaker contrastive loss (ablation) | LipsFlow · DVS-Lip | SpkVA: 78.5 -> 91.7 | Improved |
| add speaker contrastive loss (ablation) | LipsFlow · DVS-Lip | WER: 28.2 -> 26.6 | Improved |
| use 2-stage freeze-then-finetune curriculum (ablation) | LipsFlow · DVS-Lip | WER: 25.2 -> 23.8 | Improved |
| add data augmentation (MixUp, CutMix, occlusion) (ablation) | LipsFlow · DVS-Lip | WER: 23.8 -> 22.3 | Improved |
| use event-based input under degraded conditions instead of RGB | LipsFlow vs Llama-AVSR · AVA Severe Occlusion | WER: 35.1 -> 26.5 | Improved |
Closest papers by meaning
- VSRo-200: A Romanian Visual Speech Recognition Dataset for Studying Supervision and Multimodal Robustness (NeurIPS 2026 · similarity 0.86)
- ProsodyTalker: 3D Visual Speech Animation via Prosody Decomposition (AAAI 2025 · similarity 0.86)
- Learning Visual Speech Representations via Cross-Modal Distillation and Joint Face-Lip Modeling (NeurIPS 2026 · similarity 0.85)
- ES3: Evolving Self-Supervised Learning of Robust Audio-Visual Speech Representations (CVPR 2024 · similarity 0.85)
- video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models (ICML 2024 · similarity 0.85)
- From Faces to Voices: Learning Hierarchical Representations for High-quality Video-to-Speech (CVPR 2025 · similarity 0.85)
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.
