Feed-forward 3D scene reconstruction model that outputs triangle primitives directly usable as meshes for simulation.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
TriSplat is a research model for sparse-view 3D scene reconstruction. It replaces the Gaussian primitives used in most feed-forward splatting networks with oriented triangle primitives, so the output can be exported as a standard mesh without TSDF fusion or Poisson reconstruction. The project is by authors from Zhejiang University, ETH Zurich, ETH AI Center, Microsoft and Monash University (arXiv:2605.26115).
From sparse input images, a DINOv2-backed transformer decoder predicts dense local 3D point maps, relative camera poses, optional intrinsics and per-pixel triangle attributes. Triangle orientation is anchored to geometry normals, which are refined by an image-conditioned normal head. Opacity and edge-blur schedules start soft and sharpen progressively during training. At export, low-opacity triangles are dropped, face winding is corrected and nearby vertices are merged. The resulting meshes can be loaded into physics engines, collision detectors and standard renderers.
The project page offers an interactive demo with agents moving on exported mesh surfaces, plus a mesh gallery. Web versions are cropped and sampled, so they show holes that the offline exports do not have. A YouTube review states that the code is released and the model files are about 4.4 GB. No repository or weights link was available in the provided material, so those details are unverified.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
3D reconstruction model that turns images into simulation-ready scenes by representing them as triangle primitives instead of Gaussian splats, skipping mesh conversion. Models are about 4.4 GB and code is released.
Closest neighbours by meaning, not just by category.
Model
Feed-forward 3D Gaussian Splatting model that reconstructs scenes from arbitrary input views, with or without camera poses.
Research · Preferred Networks
Generative method from Preferred Networks that reconstructs high-resolution 3D scene meshes from posed multi-view images.
Model
Feed-forward system that turns an RGB image or video into editable, textured per-object 3D meshes in under a minute.
Model
Research model that decomposes a single image or scene mesh into object-level 3D meshes with consistent spatial layout.
Model
Research method that reconstructs a 4D Gaussian Splatting human from a single casual monocular video.
Research
Research method that reconstructs complete 4D (3D plus time) scenes from a single monocular video using Gaussian Splatting.
