MeshSplatting: Differentiable Rendering with Opaque Meshes

TL;DR

MeshSplatting introduces a novel mesh-based reconstruction method that jointly optimizes geometry and appearance through differentiable rendering, enabling real-time high-quality rendering compatible with AR/VR and game engines.

Contribution

It presents a new differentiable rendering approach for mesh reconstruction that improves quality and efficiency, bridging neural rendering with interactive 3D graphics.

Findings

Boosts PSNR by +0.69 dB over state-of-the-art

Trains 2x faster and uses 2x less memory

Creates smooth, high-quality meshes for real-time rendering

Abstract

Primitive-based splatting methods like 3D Gaussian Splatting have revolutionized novel view synthesis with real-time rendering. However, their point-based representations remain incompatible with mesh-based pipelines that power AR/VR and game engines. We present MeshSplatting, a mesh-based reconstruction approach that jointly optimizes geometry and appearance through differentiable rendering. By enforcing connectivity via restricted Delaunay triangulation and refining surface consistency, MeshSplatting creates end-to-end smooth, visually high-quality meshes that render efficiently in real-time 3D engines. On Mip-NeRF360, it boosts PSNR by +0.69 dB over the current state-of-the-art MiLo for mesh-based novel view synthesis, while training 2x faster and using 2x less memory, bridging neural rendering and interactive 3D graphics for seamless real-time scene interaction. The project page is available at https://meshsplatting.github.io/.