A General Pipeline for 3D Detection of Vehicles

TL;DR

This paper introduces a flexible pipeline that integrates 2D vehicle detection with 3D point cloud data to improve 3D detection accuracy in autonomous driving, adaptable to various 2D detection networks.

Contribution

It presents a novel, adaptable pipeline that combines 2D detection networks with 3D point cloud fusion and a model fitting algorithm for enhanced 3D vehicle detection.

Findings

Achieves competitive 3D detection results on KITTI dataset

Demonstrates flexibility with different 2D detection networks

Ranks second among 3D detection algorithms

Abstract

Autonomous driving requires 3D perception of vehicles and other objects in the in environment. Much of the current methods support 2D vehicle detection. This paper proposes a flexible pipeline to adopt any 2D detection network and fuse it with a 3D point cloud to generate 3D information with minimum changes of the 2D detection networks. To identify the 3D box, an effective model fitting algorithm is developed based on generalised car models and score maps. A two-stage convolutional neural network (CNN) is proposed to refine the detected 3D box. This pipeline is tested on the KITTI dataset using two different 2D detection networks. The 3D detection results based on these two networks are similar, demonstrating the flexibility of the proposed pipeline. The results rank second among the 3D detection algorithms, indicating its competencies in 3D detection.