Research map / Video / Video generation
Transformer with Controlled Attention for Synchronous Motion Captioning
Karim Radouane, Sylvie Ranwez, Julien Lagarde, Andon Tchechmedjiev
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 |
|---|---|---|---|
| apply local attention masks (D=r=10) instead of full context (ablation) | Synch Transformer · HumanML3D annotated subset | Element of: 46.98 -> 78.48 | Improved |
| apply local attention masks (D=r=10) instead of full context (ablation) | Synch Transformer · HumanML3D annotated subset | IoP: 39.96 -> 67.96 | Improved |
| apply local masks D=r=10 instead of full context (ablation) | Synch Transformer · HumanML3D | CIDEr: 69.0 -> 68.0 | No clear effect |
| apply local masks D=r=10 instead of full context (ablation) | Synch Transformer · KIT-ML | CIDEr: 132.3 -> 117.8 | Hurt |
| enlarge windows from D=r=10 to D=r=20 (ablation) | Synch Transformer · HumanML3D annotated subset | IoP: 67.96 -> 60.55 | Hurt |
| enlarge windows from D=r=10 to D=r=20 (ablation) | Synch Transformer · HumanML3D annotated subset | IoU: 46.40 -> 51.35 | Improved |
| use Transformer with controlled attention instead of MLP+GRU (ablation) | Synch Transformer · HumanML3D | CIDEr: 53.7 -> 70.3 | Improved |
| use Transformer with controlled attention instead of Spat+Adapt (ablation) | Synch Transformer · KIT-ML | CIDEr: 106.2 -> 132.3 | Improved |
| use controlled attention instead of TM2T on HumanML3D (ablation) | Synch Transformer · HumanML3D | CIDEr: 72.5 -> 70.3 | No clear effect |
Closest papers by meaning
- Language-Guided Transformer Tokenizer for Human Motion Generation (ECCV 2026 · similarity 0.89)
- LaMP: Language-Motion Pretraining for Motion Generation, Retrieval, and Captioning (ICLR 2025 · similarity 0.88)
- MoLingo: Motion-Language Alignment for Text-to-Human Motion Generation (CVPR 2026 · similarity 0.88)
- AttT2M: Text-Driven Human Motion Generation with Multi-Perspective Attention Mechanism (ICCV 2023 · similarity 0.88)
- SnapMoGen: Human Motion Generation from Expressive Texts (NeurIPS 2025 · similarity 0.88)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.87)
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.
