NeRFs & Gaussian Splats
A Gaussian splat is a highly optimised model that takes still images in a similar vein to NeRFs creating 3D positional splats that overlap each other to create a 3D model. Light, reflections and material properties are interpreted based on multiple images and angles and machine learning.
These GS models are not mesh or polygonal models and closely resemble a point cloud layout as such can be tricky to re-mesh or edit individual splat data. These Rasterised points are quicker to load and process than Realtime calculated NeRFS and are an evolution of this technology that is becoming increasingly more popular.
