Material-informed Gaussian Splatting for 3D World Reconstruction in a Digital Twin
Material-informed Gaussian Splatting for 3D World Reconstruction in a Digital Twin
This research presents a camera-only pipeline for 3D reconstruction within digital twins. The method utilizes 3D Gaussian Splatting from multi-view images, extracts semantic material masks using vision models, converts Gaussian representations to mesh surfaces with projected material labels, and assigns physics-based material properties for accurate sensor simulation. This approach aims to provide photorealistic reconstruction with physics-based material assignment while eliminating the hardware complexity and calibration requirements associated with traditional LiDAR-camera fusion methods. Validation is performed using an internal dataset from an instrumented test vehicle, leveraging LiDAR as ground truth for reflectivity validation alongside image similarity metrics.
The authors propose a camera-only pipeline for 3D reconstruction that integrates semantic material masks and physics-based properties to achieve sensor simulation fidelity comparable to LiDAR-camera fusion, validated against an internal dataset using LiDAR as ground truth for reflectivity.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
n with physics-based material assignment, providing sensor simulation fidelity comparable to LiDAR-camera fusion while eliminating hardware complexity and calibration requirements. We validate our camera-only method using an internal dataset from an instrumented test vehicle, leveraging LiDAR as ground truth for reflectivity validation alongside image similarity metrics.
Open source S4
Material-informed Gaussian Splatting for 3D World Reconstruction in a Digital Twin
Open source S1
arXiv:2511.20348v4 Announce Type: replace-cross Abstract: 3D reconstruction for Digital Twins often relies on LiDAR-based methods, which provide accurate geometry but lack the semantics and textures naturally captured by cameras. Traditional LiDAR-camera fusion approaches require complex calibration and still struggle with certain materials like glass, which are visible in images but poorly represented in point cloud
Open source S2
s. We propose a camera-only pipeline that reconstructs scenes using 3D Gaussian Splatting from multi-view images, extracts semantic material masks via vision models, converts Gaussian representations to mesh surfaces with projected material labels, and assigns physics-based material properties for accurate sensor simulation in modern graphics engines and simulators. This approach combines photorealistic reconstructio
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org