Safety-Critical Oracles for Metamorphic Testing of Deep Learning LiDAR Point Cloud Object Detectors

Replication Package for the Paper "Safety-Critical Oracles for Metamorphic Testing of Deep Learning LiDAR Point Cloud Object Detectors"The following two folders contain the code the results were generated with and exemplary LiDAR point cloud data for 14 executed test cases.PrerequisitesThe code is written in PythonThe code was run on an Nvidia GeForce RTX 2070 SUPER on Ubuntu 20.04.CodeContains the forked OpenPCDet repository where we implemented our metamorphic transformations and oracle functions. See seperated [README.md](01_Code/README.md) for further details.Experiments Data* Contains the LiDAR point cloud data from 14 executed test cases. 《面向深度学习激光雷达(LiDAR)点云目标检测器蜕变测试(Metamorphic Testing)的安全关键测试预言(Oracle)》论文复现包。以下两个文件夹分别包含生成实验结果所用的代码，以及14个已执行测试用例的示例激光雷达点云数据。前置条件：本代码基于Python编写，运行环境为Ubuntu 20.04系统搭配NVIDIA GeForce RTX 2070 SUPER显卡。代码文件夹：包含我们所分叉的OpenPCDet开源仓库，我们在此仓库中实现了蜕变测试变换逻辑与测试预言函数。详细信息请参阅独立的[README.md](01_Code/README.md)文档。实验数据文件夹：包含14个已执行测试用例对应的激光雷达点云数据。

搜集汇总

背景与挑战

背景概述

该数据集是论文'Safety-Critical Oracles for Metamorphic Testing of Deep Learning LiDAR Point Cloud Object Detectors'的复制包，主要用于支持深度学习LiDAR点云目标检测器的安全关键预言和蜕变测试研究。它包含基于OpenPCDet实现的代码（包括蜕变变换和预言函数）以及14个测试用例的示例LiDAR点云数据，适用于在Ubuntu系统上使用Python和GPU环境进行实验和验证。