MeshSplatting: Differentiable Rendering with Opaque Meshes
This work addresses the problem of integrating neural rendering with interactive 3D graphics for AR/VR and game engines, representing an incremental improvement over existing mesh-based methods.
The paper tackles the incompatibility between point-based splatting methods and mesh-based pipelines in novel view synthesis by introducing MeshSplatting, a mesh-based reconstruction approach that optimizes geometry and appearance through differentiable rendering, achieving a PSNR boost of +0.69 dB over the state-of-the-art MiLo while training 2x faster and using 2x less memory.
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/.