D4RT: Teaching AI to see the world in four dimensions

In three linesGoogle DeepMind introduces D4RT, a unified 4D reconstruction and tracking method that accelerates prior approaches up to 300x. The system captures object geometry and motion in videos with improved efficiency.

Summary generated by Claude — human-verified