Research map / 3D vision / Human pose estimation
Unsupervised 3D Pose Estimation for Hierarchical Dance Video Recognition
Xiaodan Hu, Narendra Ahuja
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 |
|---|---|---|---|
| use estimated 3D pose instead of 2D pose as input to movement recognition (averaged) (ablation) | movement LSTM · UID movement recognition | F-score: 0.93 -> 0.97 | Improved |
| use estimated 3D pose instead of 2D pose for Hips movement (ablation) | movement LSTM · UID movement recognition (Hips) | F-score: 0.81 -> 1.00 | Improved |
| use 3D pose instead of 2D pose for Head movement (ablation) | movement LSTM · UID movement recognition (Head) | F-score: 0.95 -> 0.93 | Hurt |
| use 3D pose instead of 2D pose as direct input to genre recognition (ablation) | genre LSTM · UID dance genre recognition | F-score: 0.44 -> 0.47 | Improved |
| use movements derived from 3D rather than 2D pose (ablation) | genre LSTM · UID dance genre recognition | F-score: 0.50 -> 0.55 | Improved |
| add movement features to 3D pose for genre recognition (ablation) | genre LSTM · UID dance genre recognition | F-score: 0.47 -> 0.86 | Improved |
| add movement features to 2D pose for genre recognition (ablation) | genre LSTM · UID dance genre recognition | F-score: 0.44 -> 0.73 | Improved |
| use 3D pose plus movements instead of 2D pose plus movements (ablation) | genre LSTM · UID dance genre recognition | F-score: 0.73 -> 0.86 | Improved |
| use proposed semi-supervised method instead of Pavllo semi-supervised | 3D pose estimator · AIST++ subset | MPJPE (mm): 446.1 -> 73.7 | Improved |
| use proposed semi-supervised method instead of best supervised Pavllo | 3D pose estimator · AIST++ subset | MPJPE (mm): 77.6 -> 73.7 | Improved |
| use proposed semi-supervised method instead of supervised Pavllo | 3D pose estimator · Human3.6M (S9, S11) | MPJPE (mm): 46.8 -> 47.3 | Hurt |
| use proposed unsupervised method instead of Pavllo semi-supervised with GT 2D | 3D pose estimator · AIST++ subset | MPJPE (mm): 446.1 -> 246.4 | Improved |
Closest papers by meaning
- DanceCamera3D: 3D Camera Movement Synthesis with Music and Dance (CVPR 2024 · similarity 0.89)
- BRACE: The Breakdancing Competition Dataset for Dance Motion Synthesis (ECCV 2022 · similarity 0.89)
- DanceAnyWay: Synthesizing Beat-Guided 3D Dances with Randomized Temporal Contrastive Learning (AAAI 2024 · similarity 0.88)
- DanceFix: An Exploration in Group Dance Neatness Assessment Through Fixing Abnormal Challenges of Human Pose (AAAI 2025 · similarity 0.87)
- DanceMVP: Self-Supervised Learning for Multi-Task Primitive-Based Dance Performance Assessment via Transformer Text Prompting (AAAI 2024 · similarity 0.86)
- Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition (NeurIPS 2025 · similarity 0.85)
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.
