Hybrid Transparency for Fast Particle Sorting in 3DGUT*
Session Number
1
Advisor(s)
Qi Wu, NVIDIA Spatial Intelligence Lab
Location
A147
Discipline
Computer Science
Start Date
15-4-2026 10:15 AM
End Date
15-4-2026 11:00 AM
Abstract
With just a collection of images captured of a space, 3D Gaussian Splatting (3DGS) can recreate an almost exact looking 3D representation of that space, enabling the usage of realistic 3D environments for video games and even Virtual reality. 3D Gaussian Unscented Transform (3DGUT) is an extension of 3D Gaussian Splatting (3DGS) that allows training using fisheye cameras. However, its sorting method is costly because many sorting comparisons are done to prevent incorrect particle order. Another method is to use Hybrid Transparency where it only sorts closest particles which allow shorter comparison along with the closest particles blending correctly. My project is implementing Hybrid Transparency in 3DGUT to improve rendering performance.
Hybrid Transparency for Fast Particle Sorting in 3DGUT*
A147
With just a collection of images captured of a space, 3D Gaussian Splatting (3DGS) can recreate an almost exact looking 3D representation of that space, enabling the usage of realistic 3D environments for video games and even Virtual reality. 3D Gaussian Unscented Transform (3DGUT) is an extension of 3D Gaussian Splatting (3DGS) that allows training using fisheye cameras. However, its sorting method is costly because many sorting comparisons are done to prevent incorrect particle order. Another method is to use Hybrid Transparency where it only sorts closest particles which allow shorter comparison along with the closest particles blending correctly. My project is implementing Hybrid Transparency in 3DGUT to improve rendering performance.