Research map / Video / Video generation
Recognizing Co-Speech Gestures in-the-Wild
Sindhu Hegde, K R Prajwal, Andrew Zisserman
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 long temporal motion context with cross-attention instead of the query clip alone (ablation) | semantic gesture classifier · GRW semantic classification | High-conf accuracy: 87.27 -> 93.20 | Improved |
| Extend context from 250 to 300 frames (ablation) | semantic gesture classifier · GRW semantic classification | High-conf accuracy: 93.20 -> 92.95 | No clear effect |
| Add weakly supervised pre-training before fine-tuning on clean data (ablation) | word recognition & localization · GRW word recognition | Acc@1: 16.35 -> 18.35 | Improved |
| Add weakly supervised pre-training before fine-tuning on clean data (ablation) | word recognition & localization · GRW localization | mIoU: 0.6473 -> 0.6731 | Improved |
| Train only on noisy pseudo-labeled data, without manual labels (ablation) | word recognition & localization · GRW word recognition | Acc@1: 11.50 vs 18.35 | Mixed |
| SHuBERT-based classifier vs best baseline (Gemini-3.5) | gesture classifier · GRW semantic classification | Accuracy: 67.30 -> 75.83 | Improved |
| SHuBERT-based classifier vs Gemini-3.5 on recall | gesture classifier · GRW semantic classification | Recall: 71.10 -> 69.00 | Hurt |
| Recognition and localization model vs JEGAL | word recognition & localization · AVS-Spot | Spotting accuracy: 63.6 -> 77.3 | Improved |
| Recognition and localization model vs Gemini-3.5 | word recognition & localization · GRW word recognition | Acc@5: 30.15 -> 37.30 | Improved |
| Use prompted VLM instead of trained localizer | Qwen3-VL · GRW localization | mIoU: 0.2071 vs 0.1975 random | Hurt |
Closest papers by meaning
- Understanding Co-speech Gestures in-the-wild (ICCV 2025 · similarity 0.94)
- Co$^{\mathbf{3}}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion (ICLR 2025 · similarity 0.88)
- ConvoFusion: Multi-Modal Conversational Diffusion for Co-Speech Gesture Synthesis (CVPR 2024 · similarity 0.87)
- Retrieving Semantics from the Deep: an RAG Solution for Gesture Synthesis (CVPR 2025 · similarity 0.86)
- SocialGesture: Delving into Multi-person Gesture Understanding (CVPR 2025 · similarity 0.86)
- Co-Speech Gesture Video Generation with Implicit Motion-Audio Entanglement (CVPR 2025 · similarity 0.86)
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.
