Research map / 3D vision / Human pose estimation
Gravity-Aware Monocular 3D Human-Object Reconstruction
Rishabh Dabral, Soshi Shimada, Arjun Jain, Christian Theobalt, Vladislav Golyanik
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 |
|---|---|---|---|
| add gravity-based object trajectory and human-object constraints to VNect (10 DoF) | VNect · GraviCap dataset | global root MPE (mm): 262.0 -> 224.5 | Improved |
| add gravity-based constraints to VNect (9 DoF, known focal length) | VNect · GraviCap dataset | global root MPE (mm): 262.0 -> 191.3 | Improved |
| add gravity-based constraints to VNect for bone lengths (10 DoF) | VNect · GraviCap dataset | bone length MAE (mm): 78.4 -> 63.3 | Improved |
| add gravity-based constraints to VNect for bone lengths (9 DoF) | VNect · GraviCap dataset | bone length MAE (mm): 78.4 -> 61.1 | Improved |
| add gravity constraints and mutual localisation term to VNect for root-relative pose | VNect · GraviCap dataset | root-relative MPJPE (mm): 134.0 -> 122.8 | Improved |
| add mutual localisation term and bone-length estimation to reduce jitter vs VNect | VNect · GraviCap dataset | e_smooth: 11.35 -> 10.74 | Improved |
| knowing focal length (9 DoF) versus estimating it (10 DoF) | GraviCap · GraviCap dataset | global root MPE (mm): 224.5 -> 191.3 | Improved |
| estimate focal length too (10 DoF instead of 9 DoF) | GraviCap · GraviCap dataset | 3D object position MPE (mm): 482.9 -> 445.5 | Improved |
| add Gaussian noise sigma=10 pixels to 2D object trajectory (ablation) | GraviCap · noise sensitivity | root translation error: 12.4 -> 76.3 | Hurt |
| add Gaussian noise sigma=100 pixels to 2D object trajectory (ablation) | GraviCap · noise sensitivity | root translation error: 12.4 -> >400 | Hurt |
| add Gaussian noise sigma=100 mm to 3D poses (6 DoF) (ablation) | GraviCap · noise sensitivity | root translation error: 11.7 -> 227.2 | Hurt |
| add Gaussian noise sigma=10 mm to 3D poses (10 DoF) (ablation) | GraviCap · noise sensitivity | root translation error: 26.3 -> 60.87 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- VIBE — VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GVHMR: Gravity-View Coordinates for Global Human Motion Recovery From Monocular Videos (TPAMI 2026 · similarity 0.88)
- Gravity as a Reference for Estimating a Person's Height From Video (ICCV 2019 · similarity 0.86)
- ProxyCap: Real-time Monocular Full-body Capture in World Space via Human-Centric Proxy-to-Motion Learning (CVPR 2024 · similarity 0.85)
- WHAM: Reconstructing World-grounded Humans with Accurate 3D Motion (CVPR 2024 · similarity 0.85)
- MotionPRO: Exploring the Role of Pressure in Human MoCap and Beyond (CVPR 2025 · similarity 0.84)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.83)
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.
