Research map / Video / Video generation
BRACE: The Breakdancing Competition Dataset for Dance Motion Synthesis
Davide Moltisanti, Jinyi Wu, Bo Dai, Chen Change Loy
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 |
|---|---|---|---|
| fine-tune HRNet on BRACE annotations | HRNet COCO-pretrained · BRACE pose estimation | AP: 0.158 -> 0.357 | Improved |
| fine-tune HRNet on BRACE annotations | HRNet COCO-pretrained · BRACE pose estimation | AR: 0.202 -> 0.445 | Improved |
| fine-tune HRNet on BRACE annotations | HRNet COCO-pretrained · External powermoves video | AP: 0.462 -> 0.598 | Improved |
| fine-tune HRNet on BRACE annotations | HRNet COCO-pretrained · External powermoves video | AR: 0.513 -> 0.642 | Improved |
| enlarge attention window from 15 to 25 (2 layers) (ablation) | Dance Revolution · BRACE test set | Pose FID: 0.5239 -> 1.1174 | Hurt |
| enlarge attention window from 15 to 200 (2 layers) (ablation) | Dance Revolution · BRACE test set | Pose FID: 0.5239 -> 0.7924 | Hurt |
| use 4 layers instead of 3 (window 15) (ablation) | Dance Revolution · BRACE test set | Pose FID: 0.5158 -> 0.5723 | No clear effect |
| use 3 layers instead of 2 (window 15) (ablation) | Dance Revolution · BRACE test set | Pose FID: 0.5239 -> 0.5158 | No clear effect |
| vary window size 15 to 200 (ablation) | Dance Revolution · BRACE test set | Beat alignment score: 0.269 -> 0.265 | No clear effect |
| evaluate on breakdancing instead of GT | AIST++, Dance Revolution, Dancing 2 Music · BRACE test set | Pose FID: 0.0032 -> 0.5158 | Hurt |
Closest papers by meaning
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.92)
- OpenDance: Multimodal Controllable 3D Dance Generation with Large-scale Internet Data (CVPR 2026 · similarity 0.91)
- OmniDance: Multimodal Driven Dance Video Generation with Large-scale Internet Data (ECCV 2026 · similarity 0.91)
- Walk Before You Dance: High-fidelity and Editable Dance Synthesis via Generative Masked Motion Prior (AAAI 2026 · similarity 0.90)
- DanceCamera3D: 3D Camera Movement Synthesis with Music and Dance (CVPR 2024 · similarity 0.90)
- Bidirectional Autoregessive Diffusion Model for Dance Generation (CVPR 2024 · similarity 0.89)
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.
