Long-Tailed 3D Detection via 2D Late Fusion

arXiv:2312.10986v3 Announce Type: replace
Abstract: Long-Tailed 3D Object Detection (LT3D) addresses the problem of accurately detecting objects from both common and rare classes. Contemporary multi-modal detectors achieve low AP on rare-classes (e.g., CMT only achieves 9.4 AP on stroller), presumably because training detectors end-to-end with significant class imbalance is challenging. To address this limitation, we delve into a simple late-fusion framework that ensembles independently trained uni-modal LiDAR and RGB detectors. Importantly, such a late-fusion framework allows us to leverage large-scale …

3d object 3d object detection abstract arxiv class cs.cv cs.ro detection detectors fusion low modal multi-modal object objects problem replace training type via