Research map / 3D vision / Human pose estimation
Learning to Estimate 3D Human Pose and Shape From a Single Color Image
Read the paperCheck 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 |
|---|---|---|---|
| Replace axis-angle parameter loss with rotation-matrix parameter loss (ablation) | SMPL Priors · UP-3D, Priors trained on UP-3D | mean per-vertex error (mm): 514.9 -> 140.7 | Helped |
| Replace axis-angle parameter loss with rotation-matrix parameter loss (ablation) | SMPL Priors · UP-3D, Priors trained on CMU | mean per-vertex error (mm): 589.9 -> 152.2 | Helped |
| Add per-vertex mesh loss to rotation-matrix parameter loss (ablation) | SMPL Priors · UP-3D, Priors trained on UP-3D | mean per-vertex error (mm): 140.7 -> 120.7 | Helped |
| Add per-vertex mesh loss to rotation-matrix parameter loss (ablation) | SMPL Priors · UP-3D, Priors trained on CMU | mean per-vertex error (mm): 152.2 -> 142.0 | Helped |
| Add differentiable reprojection finetuning (ablation) | SMPL Priors · UP-3D, Priors trained on UP-3D | mean per-vertex error (mm): 120.7 -> 117.7 | Helped |
| Add differentiable reprojection finetuning (ablation) | SMPL Priors · UP-3D, Priors trained on CMU | mean per-vertex error (mm): 142.0 -> 135.5 | Helped |
| Use our direct pose prediction as SMPLify initialization and anchor | SMPLify · LSP test set, foreground-background segmentation | accuracy: 91.89 -> 92.17 | Helped |
| Use our direct pose prediction as SMPLify initialization and anchor | SMPLify · LSP test set, foreground-background segmentation | F1: 88.07 -> 88.38 | Helped |
| Use our direct pose prediction as SMPLify initialization and anchor | SMPLify · LSP test set, part segmentation | accuracy: 87.71 -> 88.24 | Helped |
| Use our direct pose prediction as SMPLify initialization and anchor | SMPLify · LSP test set, part segmentation | F1: 63.98 -> 64.62 | Helped |
| Use a single direct forward pass instead of iterative SMPLify optimization | 3D pose and shape estimation · Single color image inference | runtime: 50ms; roughly 1 minute | Helped |
Closest papers by meaning
- Convolutional Mesh Regression for Single-Image Human Shape Reconstruction (CVPR 2019 · similarity 0.93)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.93)
- Probabilistic 3D Human Shape and Pose Estimation From Multiple Unconstrained Images in the Wild (CVPR 2021 · similarity 0.92)
- 3D Human Pose Estimation = 2D Pose Estimation + Matching (CVPR 2017 · similarity 0.90)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.90)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 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.