GRADE: Gaussian-Wise Rendering with Adaptive Dependency-Ordered Execution for Static and Dynamic 3DGS
Overview
GRADE schedules rendering work for static and dynamic Gaussian scenes. Its 4D implementation computes appearance only for candidates that survive filtering for the current camera. For a changing scene, the time and camera of a frame determine which primitives may contribute. The public implementation can reject records before generating geometry and defer appearance evaluation. It also includes a search over stage configurations and a benchmark for comparing execution orders.
GRADE 研究静态与动态高斯场景的渲染调度。其 4D 实现先筛选当前视角的候选高斯,再计算这些高斯的外观。动态场景中,每一帧的时间和相机位置决定哪些高斯可能参与成像。公开实现可以在生成几何信息之前排除部分高斯,推迟外观计算,并提供阶段配置搜索和执行顺序对比测试。
This work is under review. The linked resources are the currently public materials.
Research by Minnan Pei (裴旻楠) and collaborators · Email