D4RT

How DeepMind’s New AI Predicts What It Cannot See

Two Minute Papers

The single-transformer approach that handles depth, motion, and camera pose together.