Research map / Video / Video segmentation and tracking
OpenFS: Multi-Hand-Capable Fingerspelling Recognition with Implicit Signing-Hand Detection and Frame-Wise Letter-Conditioned Synthesis
Junuk Cha, Jihyeon Kim, Han-Mu Park
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 dual-level positional encoding (ablation) | OpenFS recognizer · ChicagoFSWild | letter accuracy: 73.2 -> 74.8 | Improved |
| add SF and MA auxiliary losses with standard positional encoding (ablation) | OpenFS recognizer · ChicagoFSWild | letter accuracy: 73.2 -> 73.1 | No clear effect |
| add SF and MA auxiliary losses with dual-level positional encoding (ablation) | OpenFS recognizer · ChicagoFSWild | letter accuracy: 74.8 -> 75.4 | Improved |
| replace explicit signing-hand detection with implicit detection via cross-attention | OpenFS · ChicagoFSWild | signing-hand detection accuracy: 90.4 -> 99.9 | Improved |
| add synthetic training data from generator | OpenFS · ChicagoFSWildPlus | letter accuracy: 70.5 -> 74.6 | Improved |
| add synthetic training data from generator | PoseNet · FSNeo | letter accuracy: 61.2 -> 94.9 | Improved |
| add synthetic training data from generator | OpenFS · FSNeo | letter accuracy: 80.5 -> 97.6 | Improved |
| use letter-conditioned instead of word-conditioned generation (ablation) | generator · Generated sequences (OpenFS recognizer) | letter accuracy: 23.3 -> 40.2 | Improved |
| use frame-wise letter conditioning instead of global letter conditioning (ablation) | generator · Generated sequences (OpenFS recognizer) | letter accuracy: 40.2 -> 82.3 | Improved |
| replace PoseNet CTC and re-ranking with OpenFS (no post-processing) | recognizer · Speed, batch size 1 | FPS: 6 -> 962 | Improved |
Closest papers by meaning
- Fingerspelling Detection in American Sign Language (CVPR 2021 · similarity 0.94)
- BANZ-FS: BANZSL Fingerspelling Dataset (ICLR 2026 · similarity 0.91)
- FSboard: Over 3 Million Characters of ASL Fingerspelling Collected via Smartphones (CVPR 2025 · similarity 0.91)
- Fingerspelling Recognition in the Wild With Iterative Visual Attention (ICCV 2019 · similarity 0.90)
- A Wireframe-Based Approach for Classifying and Acquiring Proficiency in the American Sign Language (Student Abstract) (AAAI 2024 · similarity 0.85)
- LERMO: A Novel Web Game for AI-Enhanced Sign Language Recognition (AAAI 2024 · similarity 0.83)
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.
