Research map / Video / Video segmentation and tracking
CoSign: Exploring Co-occurrence Signals in Skeleton-based Continuous Sign Language Recognition
Peiqi Jiao, Yuecong Min, Yanan Li, Xiaotao Wang, Lei Lei, Xilin Chen
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 group-specific GCN modules (ablation) | skeleton baseline · PHOENIX14 Dev | WER (%): 24.3 -> 22.8 | Improved |
| add group-specific centralization (ablation) | group-specific GCN · PHOENIX14 Dev | WER (%): 22.8 -> 21.8 | Improved |
| add group dropout (ablation) | CoSign-1s · PHOENIX14 Dev | WER (%): 21.8 -> 21.2 | Improved |
| add complementary regularization (ablation) | CoSign-1s with group dropout · PHOENIX14 Dev | WER (%): 21.2 -> 21.0 | Improved |
| add complementary masks (ablation) | CoSign-2s · PHOENIX14 Dev | WER (%): 20.2 -> 19.7 | Improved |
| gradual fusion instead of late fusion of skeleton and motion (ablation) | CoSign-2s · PHOENIX14 Dev | WER (%): 21.0 -> 20.7 | Improved |
| add motion stream to skeleton stream (ablation) | CoSign-1s · PHOENIX14 Dev | WER (%): 21.8 -> 20.7 | Improved |
| vary dropout mask clip length from 25 to 100 (ablation) | CoSign-1s · PHOENIX14 Dev | WER (%): 21.2 -> 21.4 | No clear effect |
| use skeleton-based CoSign-2s instead of SMKD video model | PHOENIX14 training cost · PHOENIX14 | training time: ~19.5 h -> ~4.5 h | Improved |
| use CoSign-2s on signer-independent split | PHOENIX14-SI · PHOENIX14-SI Test | WER (%): 39.0 -> 29.6 | Improved |
| use CoSign-1s instead of baseline for inference speed | inference · PHOENIX14 | inference speed (seq/s): 26.1 -> 18.8 | Hurt |
Closest papers by meaning
- HyperSign: Hierarchical Hypergraph-based Co-occurrence Modeling for Sign Language Recognition and Translation (AAAI 2026 · similarity 0.89)
- Continuous Sign Language Recognition With Correlation Network (CVPR 2023 · similarity 0.89)
- VSNet: Focusing on the Linguistic Characteristics of Sign Language (CVPR 2025 · similarity 0.89)
- TCNet: Continuous Sign Language Recognition from Trajectories and Correlated Regions (AAAI 2024 · similarity 0.88)
- OLMD: Orientation-aware Long-term Motion Decoupling for Continuous Sign Language Recognition (AAAI 2025 · similarity 0.87)
- SignGraph: A Sign Sequence is Worth Graphs of Nodes (CVPR 2024 · 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.
