Resolving Class Imbalance for LiDAR-based Object Detector by Dynamic Weight Average and Contextual Ground Truth Sampling

TL;DR

This paper introduces a novel approach for LiDAR-based 3D object detection in autonomous driving, addressing data imbalance through dynamic loss weighting and semantic-aware ground truth sampling, resulting in improved detection accuracy.

Contribution

It presents a combined method of dynamic weight averaging for multi-head detection and semantic-aware ground truth sampling to effectively mitigate class imbalance in LiDAR data.

Findings

Improved detection accuracy on KITTI and nuScenes datasets.

Effective handling of class imbalance in LiDAR-based detection.

Enhanced robustness of the detector across classes.

Abstract

An autonomous driving system requires a 3D object detector, which must perceive all present road agents reliably to navigate an environment safely. However, real-world driving datasets often suffer from the problem of data imbalance, which causes difficulties in training a model that works well across all classes, resulting in an undesired imbalanced sub-optimal performance. In this work, we propose a method to address this data imbalance problem. Our method consists of two main components: (i) a LiDAR-based 3D object detector with per-class multiple detection heads where losses from each head are modified by dynamic weight average to be balanced. (ii) Contextual ground truth (GT) sampling, where we improve conventional GT sampling techniques by leveraging semantic information to augment point cloud with sampled ground truth GT objects. Our experiment with KITTI and nuScenes datasets confirms our proposed method's effectiveness in dealing with the data imbalance problem, producing better detection accuracy compared to existing approaches.