Research map / 3D vision / Human pose estimation
Human Pose Recognition via Occlusion-Preserving Abstract Images
Saad Manzur, Wayne B Hayes
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 |
|---|---|---|---|
| enable wrap-around Gaussian heatmaps for pose and viewpoint (ablation) | abstract-to-pose network · H36M all actions | MPJPE (mm): 58.2mm -> 48.8mm | Improved |
| use seam-behind-subject wrapped viewpoint encoding instead of naive encoding (ablation) | viewpoint network · H36M | viewpoint angular error (deg): 55.44 -> 8.44 | Improved |
| increase viewpoint heatmap resolution from 16x16 to 32x32 (ablation) | viewpoint network · H36M | angular error (deg): 13.35 -> 9.24 | Improved |
| increase viewpoint heatmap resolution from 32x32 to 64x64 (ablation) | viewpoint network · H36M | angular error (deg): 9.24 -> 8.44 | Improved |
| train only on Cube variant instead of hybrid Cube+Flat (ablation) | abstract-to-pose network · H36M (Protocol #1) | MPJPE (mm): 48.8 -> 33.5 | Improved |
| use abstract images with occlusion instead of 2D keypoint stick figures for lifting | 3D pose lifting · cross-dataset (GPA, 3DPW, SURREAL) | MPJPE: 30%-40% lead in most cases | Improved |
| add more vertical camera levels in synthetic environment (ablation) | abstract-to-pose training · cross-dataset MPJPE | MPJPE (mm) | Improved |
| train on two 3D pose datasets with transformer (Goel) vs ours trained on H36M only | 3DPW cross-dataset · 3DPW | MPJPE (mm): 95.83 -> 70.0 | Hurt |
Closest papers by meaning
- Occlusion-Aware Networks for 3D Human Pose Estimation in Video (ICCV 2019 · similarity 0.92)
- 3D Human Pose Estimation Using Spatio-Temporal Networks with Explicit Occlusion Training (AAAI 2020 · similarity 0.91)
- Explicit Occlusion Reasoning for Multi-Person 3D Human Pose Estimation (ECCV 2022 · similarity 0.90)
- View-Invariant, Occlusion-Robust Probabilistic Embedding for Human Pose (IJCV 2021 · similarity 0.90)
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.88)
- Lifting by Image – Leveraging Image Cues for Accurate 3D Human Pose Estimation (AAAI 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.
