Matthias Niessner of TU Munich announces Echo-2, using generative world-modeling priors to reconstruct 3D scenes from sparse images · Digg