SpatialTracker: Tracking Any 2D Pixels in 3D Space
This addresses the challenge of pixel motion tracking in videos for computer vision applications, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the problem of recovering dense and long-range pixel motion in videos by estimating point trajectories in 3D space to mitigate issues from 2D projection, achieving state-of-the-art tracking performance, especially in challenging scenarios like out-of-plane rotation.
Recovering dense and long-range pixel motion in videos is a challenging problem. Part of the difficulty arises from the 3D-to-2D projection process, leading to occlusions and discontinuities in the 2D motion domain. While 2D motion can be intricate, we posit that the underlying 3D motion can often be simple and low-dimensional. In this work, we propose to estimate point trajectories in 3D space to mitigate the issues caused by image projection. Our method, named SpatialTracker, lifts 2D pixels to 3D using monocular depth estimators, represents the 3D content of each frame efficiently using a triplane representation, and performs iterative updates using a transformer to estimate 3D trajectories. Tracking in 3D allows us to leverage as-rigid-as-possible (ARAP) constraints while simultaneously learning a rigidity embedding that clusters pixels into different rigid parts. Extensive evaluation shows that our approach achieves state-of-the-art tracking performance both qualitatively and quantitatively, particularly in challenging scenarios such as out-of-plane rotation.