Deep Generic Dynamic Object Detection Based on Dynamic Grid Maps

TL;DR

A LiDAR-based dynamic grid is generated online and a deep learning-based detector is trained on the dynamic grid to infer the presence of dynamic objects of any type, which is a prerequisite for safe automated vehicles in arbitrary, edge-case scenarios.

Abstract

This paper describes a method to detect generic dynamic objects for automated driving. First, a LiDAR-based dynamic grid is generated online. Second, a deep learning-based detector is trained on the dynamic grid to infer the presence of dynamic objects of any type, which is a prerequisite for safe automated vehicles in arbitrary, edge-case scenarios. The Rotation-equivariant Detector (ReDet) - originally designed for oriented object detection on aerial images - was chosen due to its high detection performance. Experiments are conducted based on real sensor data and the benefits in comparison to classic dynamic cell clustering strategies are highlighted. The false positive object detection rate is strongly reduced by the proposed approach.

Figures (6)

• Figure 1: Generic dynamic object examples. Object detection methods trained for standard classes are likely to struggle in such scenarios.
• Figure 2: Moving shopping cart detection. Camera image (left) and dynamic grid with overlaid detection results (right).
• Figure 3: Precision/Recall curves of ReDet trained on different datasets. The evaluation is always implemented on the whole test set from data 1, 2, and 3.
• Figure 4: Precision/Recall curves of ReDet compared to RetinaNet.
• Figure 5: ReDet object detection precision/recall curve compared to precision/recall of classic method shown as red dot. ReDet is trained on the whole training set. Both ReDet and the classic DBSCAN are evaluated on data 1 and 3.