3D object detection is receiving increasing attention from both industry and
academia thanks to its wide applications in various fields. In this paper, we
propose Point-Voxel Region-based Convolution Neural Networks (PV-RCNNs) for 3D
object detection on point clouds. First, we propose a novel 3D detector,
PV-RCNN, which boosts the 3D detection performance by deeply integrating the
feature learning of both point-based set abstraction and voxel-based sparse
convolution through two novel steps, i.e., the voxel-to-keypoint scene encoding
and the keypoint-to-grid RoI feature abstraction. Second, we propose an
advanced framework, PV-RCNN++, for more efficient and accurate 3D object
detection. It consists of two major improvements: sectorized proposal-centric
sampling for efficiently producing more representative keypoints, and
VectorPool aggregation for better aggregating local point features with much
less resource consumption. With these two strategies, our PV-RCNN++ is about
$3\times$ faster than PV-RCNN, while also achieving better performance. The
experiments demonstrate that our proposed PV-RCNN++ framework achieves
state-of-the-art 3D detection performance on the large-scale and
highly-competitive Waymo Open Dataset with 10 FPS inference speed on the
detection range of 150m * 150m.

本文提出了一种针对点云的三维目标检测新方法 PV-RCNN，通过点和体素两个方面来深度整合特征学习，采用体素到关键点场景编码和关键点到网格 RoI 特征抽象两个创新步骤，更高效准确地实现三维目标检测。PV-RCNN++ 框架通过部分区域划分建议 - 中心采样和 VectorPool 池化等两种策略，在更低资源消耗下实现了更优秀的性能，相比 PV-RCNN 在保持准确率的同时提高了 3 倍速度，并在大规模 Waymo 开放数据集上取得了最先进的三维检测性能。

PV-RCNN++：具有局部向量表示的点 - 体素特征集抽象 用于 3D 目标检测

PV-RCNN++: Point-Voxel Feature Set Abstraction With Local Vector  Representation for 3D Object Detection

In this technical report, we present the top-performing LiDAR-only solutions
for 3D detection, 3D tracking and domain adaptation three tracks in Waymo Open
Dataset Challenges 2020. Our solutions for the competition are built upon our
recent proposed PV-RCNN 3D object detection framework. Several variants of our
PV-RCNN are explored, including temporal information incorporation, dynamic
voxelization, adaptive training sample selection, classification with RoI
features, etc. A simple model ensemble strategy with non-maximum-suppression
and box voting is adopted to generate the final results. By using only LiDAR
point cloud data, our models finally achieve the 1st place among all LiDAR-only
methods, and the 2nd place among all multi-modal methods, on the 3D Detection,
3D Tracking and Domain Adaptation three tracks of Waymo Open Dataset
Challenges. Our solutions will be available at
this https URL

本文介绍了基于 PV-RCNN 检测框架的 LiDAR 3D 检测、3D 跟踪和域自适应三项竞赛的获胜解决方案，其中探讨了 PV-RCNN 的多种变体，采用非极大值抑制和框投票的简单模型集成策略，使用 LiDAR 数据实现了 1 号位的 LiDAR-only 方法和三个轨道的 2 号位的多模态方法。