Research map / Video / Video segmentation and tracking
Transferring Cross-Domain Knowledge for Video Sign Language Recognition
Dongxu Li, Xin Yu, Chenchen Xu, Lars Petersson, Hongdong Li
Check 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 |
|---|---|---|---|
| transfer knowledge using coarse alignment and news-sign prototypes | I3D sign recognition · WLASL100 | micro top1 accuracy (%): 65.89 -> 77.52 | Improved |
| transfer knowledge using coarse alignment and news-sign prototypes | I3D sign recognition · WLASL300 | micro top1 accuracy (%): 56.14 -> 68.56 | Improved |
| transfer knowledge using coarse alignment and news-sign prototypes | I3D sign recognition · MSASL100 | micro top1 accuracy (%): 80.91 -> 83.04 | Improved |
| transfer knowledge using coarse alignment and news-sign prototypes | I3D sign recognition · MSASL200 | micro top1 accuracy (%): 74.29 -> 80.31 | Improved |
| use the proposed word recognizer as a sliding-window temporal localizer | Sign news localization · Sign news, tIoU 0.5 | mAP: 15.3 -> 28.1 | Improved |
| add coarse domain alignment through joint training on news and isolated signs (ablation) | Proposed method · WLASL100 | micro top1 accuracy (%): 70.93 -> 77.52 | Improved |
| use news-sign centroids rather than isolated-sign centroids for memory (ablation) | Coarsely aligned I3D · WLASL100 | micro top1 accuracy (%): 72.09 -> 77.52 | Improved |
| use news-sign centroids rather than isolated-sign centroids for memory (ablation) | Plain I3D · WLASL100 | micro top1 accuracy (%): 72.48 -> 77.52 | Improved |
| combine isolated-sign and news-sign features to compute memory centroids (ablation) | Coarsely aligned I3D · WLASL100 | micro top1 accuracy (%): 77.52 -> 66.67 | Hurt |
| directly add extracted news words to the training set | I3D sign recognition · WLASL100 | micro top1 accuracy (%): 65.89 -> 61.63 | Hurt |
| directly add extracted news words to the training set | I3D sign recognition · MSASL200 | micro top1 accuracy (%): 74.29 -> 75.40 | Improved |
Closest papers by meaning
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.91)
- BSL-1K: Scaling up co-articulated sign language recognition using mouthing cues (ECCV 2020 · similarity 0.90)
- CiCo: Domain-Aware Sign Language Retrieval via Cross-Lingual Contrastive Learning (CVPR 2023 · similarity 0.90)
- Read and Attend: Temporal Localisation in Sign Language Videos (CVPR 2021 · similarity 0.90)
- Automatic Dense Annotation of Large-Vocabulary Sign Language Videos (ECCV 2022 · similarity 0.90)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · 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.
