SpatialTracker is a research breakthrough presented at CVPR 2024, dedicated to recovering dense pixel motions in videos within 3D space. This method elevates 2D pixels into 3D space, utilizing a three-plane representation to express the 3D content of each frame, and iteratively updates the transformer to estimate 3D trajectories. Tracking in 3D allows us to leverage rigidity constraints while learning a rigid embedding that aggregates pixels into distinct rigid parts. Compared to other tracking methods, SpatialTracker demonstrates superior performance in both quality and quantity, especially in challenging scenarios involving out-of-plane rotations.