Presentation

<
A LoD of Gaussians: Out-of-Core Training and Rendering for Seamless Ultra-Large Scene Reconstruction
Session3D Gaussian Splatting I
DescriptionGaussian Splatting struggles with large scenes due to memory limits and chunking artifacts. We introduce "A LoD of Gaussians," a new framework enabling ultra-large scene training and rendering on a single consumer GPU. Through out-of-core streaming and Level-of-Detail, it achieves seamless multi-scale reconstruction without relying on scene partitioning.
Authors
Felix Windisch
Graz University of Technology
Thomas Köhler
Graz University of Technology
Lukas Radl
Graz University of Technology
Mattia D'urso
Graz University of Technology
Michael Steiner
Graz University of Technology
Dieter Schmalstieg
University of Stuttgart
Graz University of Technology
Markus Steinberger
Graz University of Technology
Event Type
Technical Paper
TimeMonday, 20 July 20262:40pm - 2:50pm PDT
LocationRoom 408 A
Digital Library PDF
Session Time & Location
Sunday, 19 July 20266:00pm - 8:45pm PDTHall K
Monday, 20 July 20262:00pm - 3:30pm PDTRoom 408 A
Keywords
Artificial Intelligence/Machine Learning
Geometry
Modeling
Registration Categories
Full Conference Supporter
Full Conference
Next PresentationNext PresentationMobile3DGS³: Accelerate Mobile 3DGS Rendering via Gradient-Aware Super-Sampling and Frame Interpolation
Similar Presentations
Emerging Technologies Demo - A LoD of Gaussians: Out-of-Core Training and Rendering for Seamless Ultra-Large Scene Reconstruction
ATGS: Anchored Temporal Gaussian Splatting for Long Volumetric Video Representation
Multi-feature Radiance Baking Neural Networks for Instant Volumetric Rendering