XR / Spatial Computing
A Factoring Tree That Lets You Set a Memory Budget and Train Splats to Fit It
City-scale Gaussian splatting runs out of memory before it runs out of scene. Budgeted-GS, posted October 2, proposes a factoring tree of moment-matched aggregates that gives flexible quality-to-memory tradeoffs and enables budget-centered training — which inverts the usual order of operations.