LSRM is a SOTA, feed-forward 3D reconstruction model that generates high-fidelity, relightable 3D digital twins from sparse 2D views. - View it on GitHub