Geometric-Aware Unified World Modeling
Predict future frames based on initial observation images, with optional conditions of camera trajectory actions.
Generate planning paths from pairs of observation and goal images.
Reconstruct dynamic point clouds from videos by estimating depths and camera poses.
