arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2404.15879cs.CVcs.LG
收藏

重新审视基于LiDAR的3D目标检测中的分布外检测

Revisiting Out-of-Distribution Detection in LiDAR-based 3D Object Detection

Ulm University(乌尔姆大学);Robert Bosch GmbH(罗伯特·博世有限公司);Institute of Measurement, Control, and Microtechnology(测量、控制与微技术研究所)

Michael Kösel, Marcel Schreiber, Michael Ulrich, Claudius Gläser, Klaus Dietmayer

AI总结:

针对基于LiDAR的3D目标检测中分布外目标检测研究不足的问题,提出通过扰动已知类别生成合成OOD数据,并利用预训练检测器特征训练MLP进行ID与OOD分类,同时提出无需修改点云的新评估协议,在nuScenes OOD基准上验证有效。

摘要:

基于LiDAR的3D目标检测因能够在3D中精确定位和分类目标,已成为自动驾驶的重要组成部分。然而,目标检测器在处理未知前景目标时面临严峻挑战,特别是那些未出现在其原始训练数据中的目标。这些分布外(OOD)目标可能导致误分类,对自动驾驶车辆的安全性和可靠性构成重大风险。目前,基于LiDAR的OOD目标检测尚未得到充分研究。我们通过扰动已知目标类别生成OOD目标的合成训练数据来解决这一问题。我们的思路是,与分布内(ID)目标相比,这些合成OOD目标会在目标检测器的特征图中产生不同响应。随后,我们使用预训练且固定的目标检测器提取特征,并训练一个简单的多层感知机(MLP)将每次检测分类为ID或OOD。此外,我们提出了一种新的评估协议,允许在不修改点云的情况下使用现有数据集,确保对真实场景进行更真实的评估。通过在最新提出的nuScenes OOD基准上的实验,验证了我们方法的有效性。源代码可在https://github.com/uulm-mrm/mmood3d获取。

英文摘要:

LiDAR-based 3D object detection has become an essential part of automated driving due to its ability to localize and classify objects precisely in 3D. However, object detectors face a critical challenge when dealing with unknown foreground objects, particularly those that were not present in their original training data. These out-of-distribution (OOD) objects can lead to misclassifications, posing a significant risk to the safety and reliability of automated vehicles. Currently, LiDAR-based OOD object detection has not been well studied. We address this problem by generating synthetic training data for OOD objects by perturbing known object categories. Our idea is that these synthetic OOD objects produce different responses in the feature map of an object detector compared to in-distribution (ID) objects. We then extract features using a pre-trained and fixed object detector and train a simple multilayer perceptron (MLP) to classify each detection as either ID or OOD. In addition, we propose a new evaluation protocol that allows the use of existing datasets without modifying the point cloud, ensuring a more authentic evaluation of real-world scenarios. The effectiveness of our method is validated through experiments on the newly proposed nuScenes OOD benchmark. The source code is available at https://github.com/uulm-mrm/mmood3d.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑