caoyifeng001 / PV_ENcoNet

Fast Object Detection Based on Colored Point Cloud

PV_ENcoNet

3d detection model in point cloud

Demo

Introduction

Workflow

PV_ENcoNet is an efficient multi-sensor fusion based object detection model that can be deployed on off-the-shelf edge computing device for vehicular platform. PV_ENcoNet can achieve about 17.92 and 24.25 FPS on two different edge computing platforms, and a detection accuracy comparable with the state-of-the-art models on the KITTI public dataset.

Evalution

The table lists the detailed mean Average Precision (mAP) and FPS of some models for 3D detection task. The following figure illustrates the Precise-Recall (PR) curves of these models when dealing with cars, pedestrians and cyclists. The PR curves describe the overall precision of each model.

Install

Please refer to Install.md for the installation of PV_ENcoNet.

Usage

Please refer to Get_Started.md for the usage of PV_ENcoNet.

Acknowledgments

We thanks for the opensource codebases:OpenPCDet and RandLA-Net