Research map / 3D vision / Human pose estimation
Recognizing Human Actions as the Evolution of Pose Estimation Maps
Mengyuan Liu, Junsong Yuan
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 |
|---|---|---|---|
| Add heatmap evolution to 2D pose evolution (ablation) | H1 on RGB video · NTU RGB+D cross subject | accuracy: 72.96% -> 78.80% | Improved |
| Add heatmap evolution to 2D pose evolution (ablation) | H1 on RGB video · NTU RGB+D cross view | accuracy: 77.21% -> 84.21% | Improved |
| Add heatmap evolution to 2D pose evolution (ablation) | H1 on RGB video · UTD-MHAD cross subject | accuracy: 85.63% -> 92.51% | Improved |
| Add heatmap evolution to 2D pose evolution (ablation) | H1 on RGB video · PennAction half / half | accuracy: 84.08% -> 91.39% | Improved |
| Use spatial rank pooling instead of temporal rank pooling for heatmaps (ablation) | Heatmap evolution (H3) · NTU RGB+D cross subject | accuracy: 53.91% -> 72.75% | Improved |
| Use spatial rank pooling instead of temporal rank pooling for heatmaps (ablation) | Heatmap evolution (H3) · NTU RGB+D cross view | accuracy: 54.10% -> 78.35% | Improved |
| Use spatial rank pooling instead of temporal rank pooling for heatmaps (ablation) | Heatmap evolution (H3) · UTD-MHAD cross subject | accuracy: 58.88% -> 74.88% | Improved |
| Use spatial rank pooling instead of temporal rank pooling for heatmaps (ablation) | Heatmap evolution (H3) · PennAction half / half | accuracy: 84.61% -> 87.02% | Improved |
| Add depth coordinates to 2D pose evolution (ablation) | Pose evolution from depth video · UTD-MHAD cross subject | accuracy: 85.53% -> 89.44% | Improved |
| Add depth coordinates to 2D pose evolution (ablation) | Pose evolution from depth video · NTU RGB+D cross subject | accuracy: 80.52% -> 82.38% | Improved |
| Add depth coordinates to 2D pose evolution (ablation) | Pose evolution from depth video · NTU RGB+D cross view | accuracy: 85.75% -> 86.65% | Improved |
| Add single-frame CNN features and annotated poses (ablation) | H1 + H3 on PennAction · PennAction half / half | accuracy: 91.39% -> 98.22% | Improved |
Closest papers by meaning
- Co-Evolution of Pose and Mesh for 3D Human Body Estimation from Video (ICCV 2023 · similarity 0.88)
- Synthetic Humans for Action Recognition from Unseen Viewpoints (IJCV 2021 · similarity 0.88)
- VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020 · similarity 0.88)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.87)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.86)
- MotionMap: Representing Multimodality in Human Pose Forecasting (CVPR 2025 · 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.
