Research map / 3D vision / Human pose estimation
Unite the People: Closing the Loop Between 3D and 2D Human Representations
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train the pose estimator on 91 projected landmarks rather than human-labeled 14-keypoint labels (ablation) | Pose estimator trained on UP-3D · PCK@0.2, human labels (I-P14h) | PCK@0.2: 89.11 -> 91.15 | Helped |
| Train on UP-P91 rather than UPI-P14h (ablation) | SMPLify with the paper's CNN landmarks · HumanEva | 3D joint error (mm): 81.1 -> 74.5 | Helped |
| Train on UP-P91 rather than UPI-P14h (ablation) | SMPLify with the paper's CNN landmarks · Human3.6M | 3D joint error (mm): 96.4 -> 80.7 | Helped |
| Train on UP-P14 rather than UPI-P14h (ablation) | SMPLify with the paper's CNN landmarks · HumanEva | 3D joint error (mm): 81.1 -> 79.4 | Helped |
| Train on UP-P14 rather than UPI-P14h (ablation) | SMPLify with the paper's CNN landmarks · Human3.6M | 3D joint error (mm): 96.4 -> 90.9 | Helped |
| Add silhouette segmentation to the SMPLify fitting objective (ablation) | SMPLify · LSP test | FB segmentation accuracy: 0.9176 -> 0.9217 | Helped |
| Use 91 predicted landmarks instead of 14 ground-truth keypoints for fitting (ablation) | SMPLify · LSP test | part segmentation F1: 0.5762 -> 0.6164 | Helped |
| Add global rotation optimization after direct prediction from 91 landmarks (ablation) | Direct Prediction · LSP test | part segmentation F1: 0.5102 -> 0.5304 | Helped |
| Use 91 landmarks rather than 14 landmarks for direct prediction, with rotation optimization (ablation) | Direct Prediction · LSP test | part segmentation F1: 0.5222 -> 0.5304 | Helped |
| Add FashionPose to the training data using the unified 3D representation (ablation) | Pose estimator · FashionPose test | PCK@0.2: 0.897 -> 0.9028 | Helped |
| Refit previously rejected LSP images using 91 predicted landmarks rather than 14 ground-truth keypoints (ablation) | SMPLify · LSP | accepted-fit rate: 9.3% | Helped |
Closest papers by meaning
- Probabilistic 3D Human Shape and Pose Estimation From Multiple Unconstrained Images in the Wild (CVPR 2021 · similarity 0.90)
- Accurate 3D Body Shape Regression Using Metric and Semantic Attributes (CVPR 2022 · similarity 0.90)
- Learning to Estimate 3D Human Pose and Shape From a Single Color Image (CVPR 2018 · similarity 0.90)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.89)
- MVHumanNet: A Large-scale Dataset of Multi-view Daily Dressing Human Captures (CVPR 2024 · similarity 0.88)
- GeneMAN: Generalizable Single-Image 3D Human Reconstruction from Multi-Source Human Data (NeurIPS 2025 · 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.