Research map / 3D vision / 3D generation
Text-Driven 3D Hand Motion Generation from Sign Language Data
Léore Bensabath, Mathis Petrovich, Gül Varol
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 |
|---|---|---|---|
| train on LLM-generated free-form descriptions instead of raw phonology attributes (ablation) | HandMDM · BOBSL3DT-Test seen signs | R@3: 28.78 -> 35.46 | Mixed |
| generate training text with attribute-based LLM prompts rather than gloss names only (ablation) | HandMDM · BOBSL3DT-Test seen signs | R@1: 01.00 -> 21.57 | Improved |
| add HandMotionScript descriptions to training text (ablation) | HandMDM · BOBSL3DT-Test seen signs (HMS test text) | R@1: 06.30 -> 21.68 | Improved |
| add HandMotionScript to training text (ablation) | HandMDM · ASL-Text zero-shot | R@3: 27.51 -> 35.43 | Improved |
| add HandMotionScript to training text (ablation) | HandMDM · MS-ZSSLR-W zero-shot | R@3: 24.52 -> 31.50 | Improved |
| add HandMotionScript to training text (ablation) | HandMDM · BOTH57M zero-shot | R@3: 44.19 -> 40.70 | Hurt |
| use THMR-based variant assignment instead of random (ablation) | HandMDM · BOBSL3DT-Test seen signs | R@1: 18.77 -> 21.68 | Improved |
| use THMR-based variant assignment instead of random (ablation) | HandMDM · BOBSL3DT-Test unseen signs | R@3: 25.00 -> 36.20 | Improved |
| increase training data proportion (ablation) | HandMDM · BOBSL3DT-Test seen/unseen R@1 | R@1 | Improved |
| train on BOBSL3DT instead of small target dataset | HandMDM · ASL-Text zero-shot | R@3: 16.21 -> 35.43 | Improved |
| train on BOBSL3DT instead of BOTH57M training set | HandMDM · BOTH57M zero-shot | R@1: 15.70 -> 18.61 | Improved |
Closest papers by meaning
- CLUTCH: Contextualized Language model for Unlocking Text-Conditioned Hand motion modelling in the wild (ICLR 2026 · similarity 0.91)
- MySign: A High-Fidelity Motion-Capture Dataset for 3D Sign Generation in Bahasa Isyarat Malaysia (NeurIPS 2026 · similarity 0.91)
- BOTH2Hands: Inferring 3D Hands from Both Text Prompts and Body Dynamics (CVPR 2024 · similarity 0.89)
- Hand1000: Generating Realistic Hands from Text with Only 1,000 Images (AAAI 2025 · similarity 0.89)
- Neural Sign Actors: A Diffusion Model for 3D Sign Language Production from Text (CVPR 2024 · similarity 0.88)
- Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction (CVPR 2024 · similarity 0.88)
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.
