Research map / Video / Video-language understanding
FingerCap: Fine-grained Finger-level Hand Motion Captioning
Xin Shen, Rui Zhu, Lei Shen, Zhuojie Wu, Xinyu Wang, Kaihao Zhang, Tianqing Zhu, Shuchen Wu, Chenxi Miao, Weikang Li, Deguo Xia, Jizhou Huang and 1 more
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 FiGOP pose stream and motion-aware projector over plain MM Projector SFT | Qwen3-VL-8B-Instruct · FingerCap-40K average | CIDEr: 107.28 -> 155.27 | Improved |
| add FiGOP over MM Projector SFT, HandJudge | Qwen3-VL-8B-Instruct · FingerCap-40K average | HandJudge overall: 2.13 -> 2.74 | Improved |
| fine-tune with MM Projector instead of zero-shot | Qwen3-VL-8B-Instruct · FingerCap-40K average | CIDEr: 29.14 -> 107.28 | Improved |
| fine-tune with MM Projector only, HandJudge vs Gemini | Qwen3-VL-8B-Instruct · FingerCap-40K average | HandJudge overall: 2.13 vs 2.64 (Gemini-2.5-Pro) | No clear effect |
| add FiGOP to Qwen2.5-VL-7B | Qwen2.5-VL-7B-Instruct · FingerCap-40K | CIDEr: 101.45 -> 148.20 | Improved |
| add FiGOP on out-of-distribution subsets | Qwen3-VL-8B-Instruct · OOD (BSL, HOI4D, MotionBench) | HandJudge: 1.61 -> 2.04 | Improved |
| add FiGOP on OOD, CIDEr | Qwen3-VL-8B-Instruct · OOD (BSL, HOI4D, MotionBench) | CIDEr: 47.66 -> 52.21 | Improved |
| use 8 pose frames instead of 4 (ablation) | FiGOP · FingerCap-40K average | CIDEr: 147.50 -> 155.27 | Improved |
| use 8 pose frames instead of 16 (ablation) | FiGOP · FingerCap-40K average | CIDEr: 153.80 -> 155.27 | Improved |
| train on mixed Gesture+HOI instead of Gesture only, tested on HOI (ablation) | FiGOP · FingerCap-40K HOI | CIDEr: 58.49 -> 165.31 | Improved |
| train on mixed instead of HOI only, tested on Gesture (ablation) | FiGOP · FingerCap-40K Gesture | CIDEr: 42.10 -> 146.29 | Improved |
| train on mixed instead of Gesture only, tested on Gesture (ablation) | FiGOP · FingerCap-40K Gesture | CIDEr: 154.50 -> 146.29 | Hurt |
Closest papers by meaning
- BOTH2Hands: Inferring 3D Hands from Both Text Prompts and Body Dynamics (CVPR 2024 · similarity 0.89)
- Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learning (ICLR 2025 · similarity 0.88)
- HandVQA: Diagnosing and Improving Fine-Grained Spatial Reasoning about Hands in Vision-Language Models (CVPR 2026 · similarity 0.88)
- HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing (NeurIPS 2026 · similarity 0.88)
- GigaHands: A Massive Annotated Dataset of Bimanual Hand Activities (CVPR 2025 · similarity 0.87)
- Do Egocentric Video-Language Models Truly Understand Hand-Object Interactions? (ICLR 2025 · similarity 0.87)
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.
