Title: DuoSpaceNet: Leveraging Both Bird's-Eye-View and Perspective View Representations for 3D Object Detection

Abstract: Recent advances in multi-view camera-only 3D object detection either rely on an accurate reconstruction of bird's-eye-view (BEV) 3D features or on traditional 2D perspective view (PV) image features. While both have their own pros and cons, few have found a way to stitch them together in order to benefit from "the best of both worlds". To this end, we explore a duo space (i.e., BEV and PV) 3D perception framework, in conjunction with some useful duo space fusion strategies that allow effective aggregation of the two feature representations. To the best of our knowledge, our proposed method, DuoSpaceNet, is the first to leverage two distinct feature spaces and achieves the state-of-the-art 3D object detection and BEV map segmentation results on nuScenes dataset.