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

TL;DR

This paper introduces a novel method for detecting out-of-distribution objects in LiDAR-based 3D object detection by generating synthetic OOD data, extracting features, and training a classifier, validated on a new benchmark.

Contribution

It proposes a synthetic data generation approach for OOD detection in LiDAR 3D detection and a new evaluation protocol using existing datasets.

Findings

Effective OOD detection on nuScenes benchmark

Synthetic OOD data improves detection accuracy

New evaluation protocol enhances real-world applicability

Abstract

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.