Motion-Aware Temporal Confidence Capping in ReSTIR
Evan Luo
ManuscriptReal-time RenderingPath Reuse
Evan Luo
Undergraduate Student in Computer Science
University of California Santa Barbara · Department of Computer Science
I am an undergraduate student at the Department of Computer Science, University of California Santa Barbara. My research interests include Real-time Rendering, Differentiable Rendering, and Geometry Processing.
My current research focuses on adaptive variance control for ReSTIR-based path reuse algorithms, building on the GRIS framework to derive closed-form optimal confidence caps from a bias-variance tradeoff analysis. I am also interested in geometry processing pipelines and have contributed to Blender as an open-source developer.
Research Interests
Real-time Rendering Offline Rendering Differentiable Rendering Geometry Processing
News
Recent updates and announcements
Jun 2026
Will join Huawei as a research intern working on super-resolution.
Sep 2025
Started B.Sc. in Computer Science at UC Santa Barbara.
All Publications
Complete list of publications and preprints by year
2026
Education
Academic background
B.Sc. in Computer Science
University of California Santa Barbara
2025 - 2029 (Expected)
GPA: 4.0/4.0
Experience
Research and industry positions
Research Internship
2026 (Incoming)Huawei
Hong Kong Research Center
Incoming research intern focusing on super-resolution.
AI Algorithm Internship
2026 - PresentBeijing Freedo Technology
Research Center
Develop Differentiable Physics Simulation & Differentiable Rendering pipeline for Video Generation of Fluid Dynamics with Taichi (e.g. FluidNexus-like pipeline).
Undergraduate Research Assistant
2025 - PresentUniversity of California Santa Barbara
Four Eyes Lab
Working with Prof. Tobias Höllerer and Will Zhang on Open-vocabulary Semantic Segmentation.
Projects
Open-source tools and research software
Hylux Renderer
A real-time rendering framework in C++ using Slang RHI and Slang shaders, featuring a device-host consistent Scene API inspired by Falcor, a modular RenderGraph system supporting composable RenderPasses, and a cross-language interop layer using HostFxr that enables RenderPasses to be authored in C# while the core runtime remains in C++.
C++ Rendering Slang
ReSTIR BDPT
A reproduction of ReSTIR BDPT in Falcor built solely from the paper without access to original source code, implementing bidirectional path tracing with resampled importance sampling.
Rendering Falcor Path Tracing
Blender
Contributed to the Grease Pencil module, implementing debugging layer/group traversal and UI integration across C++ core and Python RNA APIs, and investigated and fixed issues related to Grease Pencil tree structures including node ordering, parent-child relationships, and iteration correctness.
Blender 3D Modeling DCC
Contact
Feel free to reach out for collaborations or inquiries