Research map / 3D vision / Human pose estimation
VolumetricSMPL: A Neural Volumetric Body Model for Efficient Interactions, Contacts, and Collisions
Marko Mihajlovic, Siwei Zhang, Gen Li, Kaifeng Zhao, Lea Muller, Siyu Tang
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 Neural Blend Weights to base MLP (ablation) | VolumetricSMPL · PosePrior ablation | mean IoU (%): 92.75 -> 94.06 (R=1) | Improved |
| increase NBW rank R from 1 to 80 (ablation) | VolumetricSMPL · PosePrior ablation | SDF MSE x1e-5: 4.8 -> 3.7 | Improved |
| increase NBW rank R from 20 to 80 (ablation) | VolumetricSMPL · PosePrior ablation | mean IoU (%): 94.60 -> 94.67 | No clear effect |
| add low-frequency positional encoding to base MLP (ablation) | VolumetricSMPL · PosePrior ablation | SDF MSE x1e-5: 92.75 -> 93.00 IoU; 5.2 -> 8.3 SDF MSE | Mixed |
| increase decoder width from 32 to 64 neurons (ablation) | VolumetricSMPL · PosePrior ablation | mean IoU (%): 94.12 -> 94.67 | Improved |
| add analytic SDF for far-away points (ablation) | VolumetricSMPL · Efficiency | inference time (ms): 25 ms -> 15 ms; 5.0 GB -> 3.1 GB | Improved |
| use VolumetricSMPL instead of COAP | VolumetricSMPL · PosePrior body model comparison | inference time (ms): 140 -> 15 | Improved |
| use VolumetricSMPL instead of COAP | VolumetricSMPL · PosePrior body model comparison | mean IoU (%): 94.31 -> 94.67 | Improved |
| replace SDF mesh computation with VolumetricSMPL for HOI | PHOSA-style optimization · In-the-wild human-object reconstruction | collision (%): 6.33 -> 5.73 collision; 3.20 -> 3.96 cm contact | Mixed |
| replace COAP with VolumetricSMPL in EgoHMR | EgoHMR · EgoBody EgoHMR | collision: 0.191 -> 0.184 | Improved |
| replace COAP with VolumetricSMPL in motion synthesis | DartControl · Scene-constrained motion synthesis (15 scenes) | scene collision (cm): 2.78 -> 0.24 | Improved |
| replace COAP with VolumetricSMPL for self-intersection | COAP optimization · Self-intersection resolution (PROX, 16 cases) | mesh triangle collisions: 74 -> 68 | Improved |
Later papers that compared with this method
1 later accepted paper reports VolumetricSMPL as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- HuCollisionField: Resolving Self-Collisions via Neural Fields for Human Prediction (ECCV 2026 · better on 5 of 5 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- COAP — COAP: Compositional Articulated Occupancy of People (CVPR 2022) · better on 5 of 5 shared results
- LEAP — LEAP: Learning Articulated Occupancy of People (CVPR 2021) · better on 3 of 3 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
- BodyNet: Volumetric Inference of 3D Human Body Shapes (ECCV 2018 · similarity 0.89)
- NPMs: Neural Parametric Models for 3D Deformable Shapes (ICCV 2021 · similarity 0.89)
- NeuralHOFusion: Neural Volumetric Rendering Under Human-Object Interactions (CVPR 2022 · similarity 0.88)
- COAP: Compositional Articulated Occupancy of People (CVPR 2022 · similarity 0.88)
- SAM 3D Body: Robust Full-Body Human Mesh Recovery (CVPR 2026 · similarity 0.88)
- ContactField: Implicit Field Representation for Multi-Person Interaction Geometry (NeurIPS 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.
