VisuTwin Canvas
GPU-native visualization runtime for digital twins and scientific computing
An independent open-source research initiative · PlayCanvas-inspired architecture in C++23
CAST (Computer Assisted Surgical Trainer) — in-situ visualization with deterministic fixed-timestep synchronization
Developed in collaboration with the Department of ECE, University of Arizona
Visual Examples
From physically-based rendering to scientific visualization and geospatial mapping.
CAST Simulation
In-situ visualization with deterministic fixed-timestep synchronization
PBR Rendering
Physically-based materials with multi-light forward rendering
Hurricane Isabel
Multi-modal scientific visualization with isosurface extraction
Geospatial Globe
WGS84 geodesy with 3D Tiles and terrain LOD
Core Capabilities
- C++23 engine with a native Metal backend
- Fixed-timestep deterministic rendering pipeline
- Physically-based forward renderer with clustered multi-light shading
- Cascaded, variance (EVSM), and contact-hardening (PCSS) soft shadows
- Image-based lighting with reflection probes and LTC area lights
- GPU skinning, morph targets, and an animation state graph
- Gaussian splatting and GPU-simulated particle systems
- Post-processing: TAA, SSAO, depth of field, bloom, and color grading
- Composable Metal shader-chunk system with runtime overrides
- GLB/glTF loading with Draco compression on a hybrid ECS + scene graph
Research Focus
Combining real-time PBR rendering with scientific data and geospatial context typically requires stitching together multiple tools across separate processes and coordinate systems. VisuTwin Canvas aims to unify these in a single native framework.
- Digital twin visualization
- Scientific computing integration
- Geospatial rendering (planned)
- Cross-platform GPU backends (planned)