MaskBEV：面向鸟瞰图三维点云联合目标检测与足迹补全
计算机视觉与模式识别
2024-02-26 v2
摘要
近期在LiDAR点云目标检测方面的工作大多聚焦于预测物体周围的边界框。这种预测通常通过基于锚或无锚检测器预测边界框来实现，需要大量关于物体的显式先验知识才能正常工作。为弥补这些局限，我们提出MaskBEV，一种基于鸟瞰图（BEV）掩码的目标检测器神经架构。MaskBEV预测一组代表被检测物体足迹的BEV实例掩码。此外，我们的方法允许在单次前向传播中完成目标检测与足迹补全。MaskBEV还将检测问题纯粹重构为分类问题，摒弃了通常用于预测边界框的回归。我们在SemanticKITTI和KITTI数据集上评估了MaskBEV的性能，并分析了该架构的优势与局限。
引用
@article{arxiv.2307.01864,
title = {MaskBEV: Joint Object Detection and Footprint Completion for Bird's-eye View 3D Point Clouds},
author = {William Guimont-Martin and Jean-Michel Fortin and François Pomerleau and Philippe Giguère},
journal= {arXiv preprint arXiv:2307.01864},
year = {2024}
}
备注
\c{opyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works