Research map / 3D vision / Human pose estimation
Moulding Humans: Non-Parametric 3D Human Shape Estimation From Single Images
Valentin Gabeur, Jean-Sebastien Franco, Xavier Martin, Cordelia Schmid, Gregory Rogez
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 |
|---|---|---|---|
| regress depth instead of classifying 19+1 quantised depth bins (ablation) | Mould depth-map network · SURREAL | pixels with correct depth within 30mm: +5% | Improved |
| add adversarial discriminator (ablation) | Mould depth-map network · 3D HUMANS with artificial occlusions | chamfer distance error: 7% error drop | Improved |
| train on SURREAL images in addition to 3D HUMANS (ablation) | Mould depth-map network · 3D HUMANS test | depth accuracy | Hurt |
| mix SURREAL with 3D HUMANS training data (ablation) | Mould depth-map network · 3D HUMANS in realistic 3D backgrounds | depth accuracy | Improved |
| use more than 50,000 training images (ablation) | Mould depth-map network · 3D HUMANS test | depth accuracy | No clear effect |
| L1 loss instead of L2 loss (ablation) | Mould depth-map network · 3D HUMANS | depth loss | Improved |
| images containing clothing | Mould depth-map network · 3D HUMANS test subset | accuracy @50mm: 85.43% -> 83.30% | Hurt |
| images containing a manipulated object | Mould depth-map network · 3D HUMANS test subset | accuracy @50mm: 84.55% -> 79.11% | Hurt |
| predict depth maps instead of fitting SMPL to ground truth meshes | SMPL fitting baseline · 3D HUMANS | depth accuracy | Improved |
Closest papers by meaning
- Learning to Estimate 3D Human Pose and Shape From a Single Color Image (CVPR 2018 · similarity 0.91)
- COSMU: Complete 3D human shape from monocular unconstrained images (ECCV 2024 · similarity 0.90)
- Probabilistic 3D Human Shape and Pose Estimation From Multiple Unconstrained Images in the Wild (CVPR 2021 · similarity 0.89)
- Learning From Synthetic Humans (CVPR 2017 · similarity 0.89)
- End-to-End Recovery of Human Shape and Pose (CVPR 2018 · similarity 0.89)
- Refining 3D Human Texture Estimation From a Single Image (TPAMI 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.
