USC: Uncompromising Spatial Constraints for Safety-Oriented 3D Object Detectors in Autonomous Driving

arXiv:2209.10368v4 Announce Type: replace-cross
Abstract: We consider the safety-oriented performance of 3D object detectors in autonomous driving contexts. Specifically, despite impressive results shown by the mass literature, developers often find it hard to ensure the safe deployment of these learning-based perception models. Attributing the challenge to the lack of safety-oriented metrics, we hereby present uncompromising spatial constraints (USC), which characterize a simple yet important localization requirement demanding the predictions to fully cover the objects when seen from the autonomous vehicle. …

3d object abstract arxiv autonomous autonomous driving challenge constraints cs.ai cs.cv cs.lg deployment detectors developers driving literature object perception performance results safe safety spatial type usc