Inconsistency-based Active Learning for LiDAR Object Detection

TL;DR

This paper introduces inconsistency-based active learning strategies for LiDAR object detection, significantly reducing labeling efforts while maintaining high detection accuracy in autonomous driving datasets.

Contribution

It extends active learning to the LiDAR domain with novel inconsistency-based sample selection methods, demonstrating substantial data efficiency improvements.

Findings

Naive inconsistency approach matches random sampling performance with 50% less labeled data.

Inconsistency strategies reduce labeling costs while maintaining detection accuracy.

Effective in various autonomous driving dataset settings.

Abstract

Deep learning models for object detection in autonomous driving have recently achieved impressive performance gains and are already being deployed in vehicles worldwide. However, current models require increasingly large datasets for training. Acquiring and labeling such data is costly, necessitating the development of new strategies to optimize this process. Active learning is a promising approach that has been extensively researched in the image domain. In our work, we extend this concept to the LiDAR domain by developing several inconsistency-based sample selection strategies and evaluate their effectiveness in various settings. Our results show that using a naive inconsistency approach based on the number of detected boxes, we achieve the same mAP as the random sampling strategy with 50% of the labeled data.