A real-time 3D rendering technique that represents a scene as a cloud of 3D Gaussian ellipsoids (blobs) rather than a continuous volumetric field. Unlike NeRFs which use neural networks to query color and density at every point, 3DGS explicitly stores parameters for each Gaussian (position, rotation, scale, opacity, and spherical harmonic coefficients for color) and uses a fast rasterization pipeline to project them onto the screen. This allows for NeRF-quality visuals with much faster training and real-time rendering speeds.
Introduced in the SIGGRAPH 2023 paper '3D Gaussian Splatting for Real-Time Radiance Field Rendering' by Kerbl et al. from Inria and MPI.
Quickly adopted by the graphics community and industry for digital twins, VR, and game development due to its speed and high fidelity.