Research map / Video / Video-language understanding
Auslan-Daily: Australian Sign Language Translation for Daily Communication and News
Xin Shen, Shaozu Yuan, Hongwei Sheng, Heming Du, 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 |
|---|---|---|---|
| feed multi-person clips without cropping signers | GASLT (RGB) · Auslan-Daily Communication | BLEU-4: 4.73 -> 4.23 | Hurt |
| signer detection before translation (SD+SLT, pose) (ablation) | GASLT/SLT models · Auslan-Daily Communication multi-person | BLEU-4: 4.23 -> 8.38 | Improved |
| signer detection before translation (SD+SLT, pose) (ablation) | SLT models · Auslan-Daily News multi-person | BLEU-4: 2.26 -> 2.53 | Improved |
| pose input instead of RGB (ablation) | GASLT · Auslan-Daily Communication single-person | BLEU-4: 4.73 -> 9.95 | Improved |
| pose input instead of RGB (ablation) | SL-Luong · Auslan-Daily Communication single-person | BLEU-4: 4.66 -> 9.45 | Improved |
| pose input instead of RGB (ablation) | GASLT · Auslan-Daily News single-person | BLEU-4: 2.56 -> 2.72 | No clear effect |
| fine-tune alignment model on Mixed instead of Communication only (test on Comm.) (ablation) | SAT · Sign Language Alignment | F1@.50: 66.33 -> 64.27 | Hurt |
| fine-tune on Mixed instead of News only (test on News) (ablation) | SAT · Sign Language Alignment | F1@.50: 50.08 -> 53.05 | Improved |
| fine-tune pre-trained SAT on Auslan | SAT · Sign Language Alignment | F1 | Improved |
| pose-based TGCN instead of RGB I3D (ablation) | TGCN · Isolated Sign Language Recognition | Top-1: 11.87 -> 14.82 | Improved |
| pose-based TGCN instead of RGB I3D (ablation) | TGCN · Signer Detection | Top-1: 89.01 -> 88.35 | No clear effect |
Closest papers by meaning
- MM-WLAuslan: Multi-View Multi-Modal Word-Level Australian Sign Language Recognition Dataset (NeurIPS 2024 · similarity 0.91)
- Scaling Sign Language Translation (NeurIPS 2024 · similarity 0.89)
- YouTube-SL-25: A Large-Scale, Open-Domain Multilingual Sign Language Parallel Corpus (ICLR 2025 · similarity 0.88)
- Lost in Translation, Found in Embeddings: Sign Language Translation and Alignment (NeurIPS 2026 · similarity 0.88)
- Lost in Translation, Found in Context: Sign Language Translation with Contextual Cues (CVPR 2025 · similarity 0.87)
- LLMs are Good Sign Language Translators (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.
