Research map / 3D vision / Gaussian splatting
SignNet-1M: Large-Scale Multilingual Sign Language Video Dataset with Downstream Benchmarks
Zhewen He, Junyi Hu, Haomian Huang, Zhenhua Li, Yushen Liu, Yi Fang
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 |
|---|---|---|---|
| train on SignNet-1M instead of original data, evaluated on SignNet-1M test | SpaMo · Phoenix14T SLT | BLEU-4: 7.81 -> 18.98 | Improved |
| train on SignNet-1M, evaluated on SignNet-1M test | UniSign · OpenASL SLT | BLEU-4: 8.12 -> 22.83 | Improved |
| train on SignNet-1M, evaluated on original test set | SpaMo · How2Sign SLT | BLEU-4: 10.11 -> 18.48 | Improved |
| train on SignNet-1M, evaluated on original test set | SpaMo · CSL-Daily SLT | BLEU-4: 20.55 -> 20.93 | No clear effect |
| train on SignNet-1M, evaluated on original test set | SpaMo · Phoenix14T SLT | BLEU-4: 22.49 -> 27.78 | Improved |
| train on SignNet-1M, evaluated on SignNet-1M test (UniSign vs original) | UniSign · CSL-Daily SLT | BLEU-4: 25.27 -> 21.66 | Mixed |
| train on SignNet-1M instead of original data | Online-CSLR · Phoenix14T SLR | WER: 51.45% -> 28.19% | Improved |
| train on SignNet-1M instead of original data | UniSign · CSL-Daily SLR | WER: 56.90% -> 30.74% | Improved |
| use identity editing alone instead of scene editing alone (ablation) | SpaMo · Phoenix14T SLT | BLEU-4: 12.53 -> 22.04 (trained) | Improved |
| add dynamic camera augmentation to novel-view (ablation) | SpaMo · Phoenix14T SLT | BLEU-4: 18.98 -> 17.13 (trained) | Hurt |
| add zoom augmentation (ablation) | SpaMo · Phoenix14T SLT | BLEU-4: 7.10 -> 8.89 (zero-shot) | Improved |
| add pitch augmentation to novel-view (ablation) | SpaMo · How2Sign SLT | BLEU-4: 6.09 -> 4.06 (zero-shot) | Hurt |
Closest papers by meaning
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.92)
- Scaling Sign Language Translation (NeurIPS 2024 · similarity 0.92)
- YouTube-SL-25: A Large-Scale, Open-Domain Multilingual Sign Language Parallel Corpus (ICLR 2025 · similarity 0.92)
- LLMs are Good Sign Language Translators (CVPR 2024 · similarity 0.90)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.90)
- Conditional Variational Autoencoder for Sign Language Translation with Cross-Modal Alignment (AAAI 2024 · similarity 0.90)
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.
