Research method that reconstructs 3D surfaces from a variable number of unposed images using flow matching.
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.
Surflo is a research project for reconstructing 3D surfaces from images using flow matching. It takes a variable number of unposed images, meaning no known camera poses, and fuses them into a single global latent state of a fixed size (K=128 tokens) from which the surface is generated.
According to the project page and a YouTube creator summary, the approach turns multiple unaligned images into a clean 3D model through this shared global state. It is aimed at researchers working on multi-view reconstruction without pose estimation.
As of the information available, no code, model weights or paper link were found on the project page, and a creator noted the code had not yet been released. Authorship and vendor are not stated in the material provided.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Turns multiple unaligned images without known camera poses into a clean 3D model via one shared global state. Code not yet released.
Closest neighbours by meaning, not just by category.
Model
Research model that builds complete 3D assets from sparse, unposed images by jointly doing reconstruction and generation.
Research
Post-training method that refines feed-forward 3D reconstruction models (Pi3, DA3, VGGT) on unlabeled video.
Model
3D reconstruction model that recovers geometry and camera poses from photos or video using a loop transformer.
Research
Multi-view video diffusion model that turns monocular or sparse videos of dynamic humans into dense multi-view videos for 4D Gaussian splats.
Model
Feed-forward 3D scene reconstruction model that outputs triangle primitives directly usable as meshes for simulation.
Model
Research method that recovers 6-DoF object and camera pose from monocular video via proxy-video translation and PnP.