Triangle Splatting+: Differentiable Rendering with Opaque Triangles
Jan Held, Renaud Vandeghen, Sanghyun Son, Daniel Rebain, Matheus Gadelha, Yi Zhou, Ming C. Lin, Marc Van Droogenbroeck, Andrea Tagliasacchi
2025-10-06
Summary
This paper introduces a new technique called Triangle Splatting+ for creating 3D models from images, focusing on making those models directly usable in virtual reality and other graphics applications.
What's the problem?
Recent methods like 3D Gaussian Splatting are great for quickly creating realistic 3D scenes, but they represent objects as a bunch of blurry blobs (Gaussians). These blobs don't work well with traditional 3D tools and game engines that rely on triangles, the building blocks of most 3D graphics. Converting these blobs into triangles usually takes extra steps, makes things more complicated, and reduces the quality of the final model.
What's the solution?
Triangle Splatting+ skips the blob step altogether and directly creates 3D models using triangles from the start. It's designed to connect these triangles together smoothly and ensure they are solid, meaning light doesn't pass through them. This creates a standard 3D mesh that can be immediately used in any graphics program without needing extra conversion steps.
Why it matters?
This research is important because it allows for faster and higher-quality 3D models that are immediately compatible with existing 3D technology like VR headsets and game engines. This opens the door to more realistic and interactive 3D experiences, and allows for further processing like physics simulations or virtual walkthroughs that weren't easily possible with previous methods.
Abstract
Reconstructing 3D scenes and synthesizing novel views has seen rapid progress in recent years. Neural Radiance Fields demonstrated that continuous volumetric radiance fields can achieve high-quality image synthesis, but their long training and rendering times limit practicality. 3D Gaussian Splatting (3DGS) addressed these issues by representing scenes with millions of Gaussians, enabling real-time rendering and fast optimization. However, Gaussian primitives are not natively compatible with the mesh-based pipelines used in VR headsets, and real-time graphics applications. Existing solutions attempt to convert Gaussians into meshes through post-processing or two-stage pipelines, which increases complexity and degrades visual quality. In this work, we introduce Triangle Splatting+, which directly optimizes triangles, the fundamental primitive of computer graphics, within a differentiable splatting framework. We formulate triangle parametrization to enable connectivity through shared vertices, and we design a training strategy that enforces opaque triangles. The final output is immediately usable in standard graphics engines without post-processing. Experiments on the Mip-NeRF360 and Tanks & Temples datasets show that Triangle Splatting+achieves state-of-the-art performance in mesh-based novel view synthesis. Our method surpasses prior splatting approaches in visual fidelity while remaining efficient and fast to training. Moreover, the resulting semi-connected meshes support downstream applications such as physics-based simulation or interactive walkthroughs. The project page is https://trianglesplatting2.github.io/trianglesplatting2/.