mitsuba-renderer/mitsuba3

What it solves

Mitsuba 3 provides a research-oriented system for simulating both forward and inverse light transport. It addresses the need for a flexible, high-performance renderer that can be used for computer graphics research and other disciplines requiring precise light simulation.

How it works

The system uses a core library and a set of plugins for materials, light sources, and rendering algorithms. It is built on Dr.Jit, a specialized just-in-time (JIT) compiler that allows the same code to be retargeted for different tasks, such as simulating scalar RGB transport or differential spectral transport on a GPU. It utilizes an LLVM backend for CPUs and a CUDA/OptiX backend for NVIDIA GPUs to achieve high performance.

Who it’s for

It is designed for researchers in computer graphics and related fields who need a differentiable renderer to compute derivatives of simulations with respect to input parameters like camera pose, geometry, and textures.

Highlights

• Differentiable Rendering: Can compute derivatives of the entire simulation relative to input parameters.
• Python-First Integration: Allows materials, textures, and algorithms to be developed in Python and JIT-compiled on the fly.
• Retargetable Architecture: Supports multiple variants including monochromatic, RGB, and spectral rendering, with optional polarization.
• Hardware Acceleration: Leverages NVIDIA GPUs with ray tracing hardware via CUDA/OptiX.