Research map / 3D vision / 3D generation
GaussianProperty: Integrating Physical Properties to 3D Gaussians with LMMs
Xinli Xu, Wenhang Ge, Dicong Qiu, ZhiFei Chen, Dongyu Yan, Zhuoyun Liu, Haoyu Zhao, Hanfeng Zhao, Shunsi Zhang, Junwei Liang, Ying-Cong Chen
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 global-to-local knowledge integration (ablation) | GaussianProperty · ABO material segmentation | mIoU (%): 22.17 -> 55.83 | Improved |
| add global-to-local knowledge integration (no voting) (ablation) | GaussianProperty · ABO material segmentation | mIoU (%): 22.17 -> 51.28 | Improved |
| add frequency-based voting (ablation) | GaussianProperty · ABO material segmentation | mIoU (%): 51.28 -> 55.83 | Improved |
| replace NeRF2Physics with GaussianProperty | NeRF2Physics · ABO material segmentation | mIoU (%): 25.59 -> 55.83 | Improved |
| replace NeRF2Physics with GaussianProperty | NeRF2Physics · MVImgNet material segmentation | mIoU (%): 4.02 -> 34.83 | Improved |
| use adaptive grasping force from estimated properties instead of fixed maximum force | MaxGF · Robot grasping (16 objects) | Success Rate (%): 75.00 -> 100.00 | Improved |
| use adaptive grasping force instead of fixed minimum force | MinGF · Robot grasping (16 objects) | Success Rate (%): 50.00 -> 100.00 | Improved |
| use adaptive grasping force instead of fixed medium force | MidGF · Robot grasping (16 objects) | No-damage Rate (%): 81.25 -> 100.00 | Improved |
| replace NeRF2Physics with GaussianProperty for hardness | NeRF2Physics · Shore hardness estimation | ADE: 35.917 -> 28.583 | Improved |
| replace NeRF2Physics with GaussianProperty for mass | NeRF2Physics · ABO mass estimation | ADE: 12.761 -> 5.960 | Improved |
| replace NeRF2Physics with GaussianProperty for mass | NeRF2Physics · ABO mass estimation | APE: 0.589 -> 1.609 | Hurt |
Later papers that compared with this method
1 later accepted paper reports GaussianProperty as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- SiPhy: Single-Image Physical Property Reasoning (ECCV 2026 · better on 2 of 5 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
- PhysGS: Bayesian-Inferred Gaussian Splatting for Physical Property Estimation (CVPR 2026 · similarity 0.90)
- GIC: Gaussian-Informed Continuum for Physical Property Identification and Simulation (NeurIPS 2024 · similarity 0.89)
- OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation (ICLR 2025 · similarity 0.88)
- Physical Property Understanding from Language-Embedded Feature Fields (CVPR 2024 · similarity 0.87)
- PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamics (CVPR 2024 · similarity 0.86)
- MonoPhysics: Estimating Geometry, Appearance, and Physical Parameters from Monocular Videos (NeurIPS 2026 · similarity 0.85)
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.
