Adobe's Large Reconstruction Model can generate 3D models from 2D images in 5 seconds, representing a major advance in 3D reconstruction.

A neural field network can create a continuous 3D model from a limited number of 2D images, and it does it without being trained on other samples.