NEW APPLE AI MODEL GENERATES 3D SCENES FROM JUST THREE IMAGES

Summary from the TLDR Design newsletter

Apple's Machine Learning team built Matrix3D, a photogrammetry model that makes detailed 3D reconstructions from only three 2D images using a unified process. It uses masked learning like early Transformer AI, training effectively on smaller datasets. Aimed at devices like Apple Vision Pro, Matrix3D is open source on GitHub with sample videos and point cloud demos online.