SpOT: Spatiotemporal Modeling for 3D Object Tracking

3D multi-object tracking aims to uniquely and consistently identify all mobile entities through time. Despite the rich spatiotemporal information available in this setting, current methods primarily rely on abstracted limited history, e.g. single-frame object bounding boxes. In work, we develop a holistic representation of traffic scenes that leverages both spatial temporal actors scene. Specifically, reformulate as problem by representing tracked objects sequences time-stamped points boxes over long history. At each timestamp, improve location motion estimates our learned refinement full sequence By considering time space jointly, naturally encodes fundamental physical priors such permanence consistency across Our framework achieves state-of-the-art performance Waymo nuScenes benchmarks.