Research map / 3D vision / 3D generation
Cross-View Isolated Sign Language Recognition via View Synthesis and Feature Disentanglement
Xin Shen, Xinyu Wang, Lei Shen, Kaihao Zhang, Xin Yu
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 view classification CE loss (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 34.36 -> 36.54 | Improved |
| add sign contrastive loss (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 36.54 -> 40.34 | Improved |
| add view contrastive loss (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 40.34 -> 40.72 | Improved |
| add full multi-task contrastive losses (ablation) | DSTA-SLR backbone · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 33.49 -> 40.25 | Improved |
| use 64 instead of 32 semantic tokens (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 38.49 -> 40.72 | Improved |
| increase semantic tokens from 64 to 128 (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 40.72 -> 28.17 | Hurt |
| add viewpoint perturbation (±10°, ±15°) to 7 common views (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 37.57 -> 40.72 | Improved |
| use larger perturbation (±15°, ±20°) instead of (±10°, ±15°) (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 40.72 -> 38.92 | Hurt |
| use 7 rather than 3 common views (ablation) | CMVSR · MM-WLAuslan AVG non-frontal | Top-1 accuracy: 37.28 -> 40.72 | Improved |
| add random clipping (ablation) | CMVSR · MTV-Test | Top-1 accuracy: 27.88 -> 28.80 | Improved |
| add speed variation (ablation) | CMVSR · MTV-Test | Top-1 accuracy: 27.88 -> 29.75 | Improved |
| add both random clipping and speed variation (ablation) | CMVSR · MTV-Test | Top-1 accuracy: 27.88 -> 31.14 | Improved |
Closest papers by meaning
- CVT-SLR: Contrastive Visual-Textual Transformation for Sign Language Recognition With Variational Alignment (CVPR 2023 · similarity 0.87)
- Improving Continuous Sign Language Recognition with Cross-Lingual Signs (ICCV 2023 · similarity 0.86)
- MM-WLAuslan: Multi-View Multi-Modal Word-Level Australian Sign Language Recognition Dataset (NeurIPS 2024 · similarity 0.86)
- VSNet: Focusing on the Linguistic Characteristics of Sign Language (CVPR 2025 · similarity 0.84)
- SignGraph: A Sign Sequence is Worth Graphs of Nodes (CVPR 2024 · similarity 0.84)
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.84)
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.
