Revisiting 3D Object Detection From an Egocentric Perspective

3D object detection is a key module in safety-critical robotics applications such as autonomous driving. For such applications, we care the most about how the detections impact the ego-agent's behavior and safety (the egocentric perspective). Intuitively, we seek more accurate descriptions of object geometry when it's more likely to interfere with the ego-agent's motion trajectory. However, current detection metrics, based on box Intersection-over-Union (IoU), are object-centric and are not designed to capture the spatio-temporal relationship between objects and the ego-agent. To address this issue, we propose a new egocentric measure to evaluate 3D object detection: Support Distance Error (SDE). Our analysis based on SDE reveals that the egocentric detection quality is bounded by the coarse geometry of the bounding boxes. Given the insight that SDE can be improved by more accurate geometry descriptions, we propose to represent objects as amodal contours, specifically amodal star-shaped polygons, and devise a simple model, StarPoly, to predict such contours. Our experiments on the large-scale Waymo Open Dataset show that SDE better reflects the impact of detection quality on the ego-agent's safety compared to IoU; and the estimated contours from StarPoly consistently improve the egocentric detection quality over recent 3D object detectors. Recent works have introduced a few modifications to evaluation protocols to address these issues, e.g., breaking down the metrics into different distance buckets [53] or using learned planning models to reflect detection quality [34] . However, they are either very coarse [53] or rely on optimized neural networks [34] , making it difficult to interpret and compare results in different settings. In this