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

Published on: April 24, 2024

Impact Score: 7.2

Arxiv code: Arxiv:2404.15879

Summary

• What is new: Addressing the challenge of detecting out-of-distribution (OOD) objects in LiDAR-based 3D object detection by generating synthetic training data.
• Why this is important: LiDAR-based object detectors struggle with unknown foreground objects not present in their training data, leading to potential safety risks.
• What the research proposes: Generating synthetic OOD objects and using a pre-trained object detector with an MLP to differentiate between ID and OOD objects.
• Results: Successfully validated the method on the nuScenes OOD benchmark, showing improved detection of OOD objects in 3D.

Technical Details

Technological frameworks used: Pre-trained object detector, synthetic data generation

Models used: Multilayer Perceptron (MLP)

Data used: nuScenes dataset

Potential Impact

Automotive companies and autonomous driving technology providers; could impact safety feature development in autonomous vehicles.

Want to implement this idea in a business?

We have generated a startup concept here: LidarGuard.