Distribution Discrepancy and Feature Heterogeneity for Active 3D Object Detection

TL;DR

This paper introduces DDFH, a novel active learning method for LiDAR-based 3D object detection that reduces annotation costs by assessing distribution discrepancy and feature heterogeneity.

Contribution

The paper proposes DDFH, a new active learning approach that considers geometric and embedding features at multiple levels to improve data efficiency in 3D detection.

Findings

DDFH reduces annotation costs by 56.3% on KITTI and Waymo datasets.

DDFH outperforms state-of-the-art methods in active 3D object detection.

DDFH is effective with both one-stage and two-stage detection models.

Abstract

LiDAR-based 3D object detection is a critical technology for the development of autonomous driving and robotics. However, the high cost of data annotation limits its advancement. We propose a novel and effective active learning (AL) method called Distribution Discrepancy and Feature Heterogeneity (DDFH), which simultaneously considers geometric features and model embeddings, assessing information from both the instance-level and frame-level perspectives. Distribution Discrepancy evaluates the difference and novelty of instances within the unlabeled and labeled distributions, enabling the model to learn efficiently with limited data. Feature Heterogeneity ensures the heterogeneity of intra-frame instance features, maintaining feature diversity while avoiding redundant or similar instances, thus minimizing annotation costs. Finally, multiple indicators are efficiently aggregated using Quantile Transform, providing a unified measure of informativeness. Extensive experiments demonstrate that DDFH outperforms the current state-of-the-art (SOTA) methods on the KITTI and Waymo datasets, effectively reducing the bounding box annotation cost by 56.3% and showing robustness when working with both one-stage and two-stage models.