E-RayZer: Self-supervised 3D Reconstruction as Spatial Visual Pre-training

E-RayZer, a self-supervised 3D Vision model predicting camera poses and scene geometry as 3D Gaussians.

Upload your images above or pick a curated example below.