Gaussian Splatting
Gaussian Splatting is an advanced 3D rendering method used to reconstruct and display real-world or synthetic environments by representing them as a set of anisotropic 3D Gaussians, mathematical functions with a bell-curve shape in space. Each “splat” is defined by its spatial position, covariance (defining size and orientation), colour (RGB), opacity (alpha), and other visual parameters.
Unlike traditional 3D graphics techniques that rely on polygon meshes or point clouds, Gaussian Splatting uses continuous volumetric primitives that can overlap, creating smooth, photorealistic representations of complex surfaces without explicit connectivity or topology. Because the Gaussians are differentiable and can be rendered analytically, the method supports real-time rendering, global illumination effects, and high-fidelity reproduction of lighting phenomena such as reflection and refraction.
Gaussian Splatting builds on principles from neural rendering and implicit representations. It allows the rendering of large-scale environments with high efficiency, making it suitable for digital twins, VR/AR applications, scanning-based reconstruction, and interactive visualisation.
From a technical point of view, this technique excels in scenarios where precise surface reconstruction is difficult or where traditional triangle meshes result in artefacts or excessive memory usage. Gaussian Splatting is compatible with various forms of input data, such as LiDAR point clouds, photogrammetry outputs, or RGB-D images, and it dynamically adjusts resolution to balance visual fidelity and performance.
Company
Integrating physical asset management with dynamic true-to-life 3D facility visualisation.
Your tursted companion in industrial planning and for the whole life cycle of your industrial facility.