TriSplat: Simulation-Ready Feed-Forward 3D Scene Reconstruction
2026-05-25 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present TriSplat, a new method that reconstructs 3D scenes from a few images by predicting triangles directly, rather than using Gaussian shapes like before. Their approach produces ready-to-use 3D meshes in one step, without extra processing, which is especially useful when camera positions are unknown. They improve the accuracy of triangle orientations using predicted 3D points and image information, making the final results more faithful to real geometry. Experiments show their method creates better shapes and still renders new views well, and the output can be used immediately in simulations and physics engines.
3D reconstructionfeed-forward networkstriangle primitivesmesh extractioncamera pose estimationsurface normalsRealEstate10KDL3DVrenderingsimulation-ready mesh
Authors
Weijie Wang, Zimu Li, Jinchuan Shi, Zeyu Zhang, Botao Ye, Marc Pollefeys, Donny Y. Chen, Bohan Zhuang
Abstract
Sparse-view 3D reconstruction is increasingly addressed with feed-forward splatting networks that predict explicit primitives directly from images. Yet most existing methods remain centered on Gaussian primitives and expose surfaces only indirectly: extracting a usable mesh for downstream simulation, physics reasoning, or embodied interaction still requires expensive post-hoc steps that break the feed-forward promise. This limitation is especially pronounced in pose-free settings, where scene structure and camera parameters must be estimated jointly from sparse observations. We present TriSplat, a feed-forward reconstruction network that represents scenes with oriented triangle primitives and directly exports simulation-ready mesh scenes from a single forward pass. Given input images, the network predicts local 3D point maps, triangle attributes, camera poses, and optional intrinsics. Rather than regressing triangle orientation as an unconstrained latent variable, our approach constructs geometry normals from the predicted point maps, refines them with an image-conditioned normal head, and converts them into stable local frames for triangle parameterization. A mono-normal bootstrap schedule further stabilizes early training, while opacity and blur scheduling progressively sharpens the learned surface representation for direct mesh extraction. Experiments on RealEstate10K and DL3DV show that this representation produces more geometry-faithful reconstructions than Gaussian feed-forward baselines while maintaining competitive novel-view rendering quality. Because the rendering primitives are themselves surface triangles, the output can be directly ingested by physics engines, collision detectors, and standard rendering pipelines without any conversion, making it a practical simulation-ready solution for feed-forward 3D scene reconstruction.