Rethinking IoU-based Optimization for Single-stage 3D Object Detection

TL;DR

This paper introduces RDIoU, a novel rotation-decoupled IoU metric that improves the training efficiency and accuracy of single-stage 3D object detectors by addressing rotation sensitivity issues.

Contribution

It proposes RDIoU, a new IoU-based optimization method that decouples rotation, enhancing training stability and detection performance in 3D object detection.

Findings

RDIoU improves detection accuracy on KITTI and Waymo datasets.

It reduces training instability caused by rotation sensitivity.

The method enhances bounding box regression precision.

Abstract

Since Intersection-over-Union (IoU) based optimization maintains the consistency of the final IoU prediction metric and losses, it has been widely used in both regression and classification branches of single-stage 2D object detectors. Recently, several 3D object detection methods adopt IoU-based optimization and directly replace the 2D IoU with 3D IoU. However, such a direct computation in 3D is very costly due to the complex implementation and inefficient backward operations. Moreover, 3D IoU-based optimization is sub-optimal as it is sensitive to rotation and thus can cause training instability and detection performance deterioration. In this paper, we propose a novel Rotation-Decoupled IoU (RDIoU) method that can mitigate the rotation-sensitivity issue, and produce more efficient optimization objectives compared with 3D IoU during the training stage. Specifically, our RDIoU simplifies the complex interactions of regression parameters by decoupling the rotation variable as an independent term, yet preserving the geometry of 3D IoU. By incorporating RDIoU into both the regression and classification branches, the network is encouraged to learn more precise bounding boxes and concurrently overcome the misalignment issue between classification and regression. Extensive experiments on the benchmark KITTI and Waymo Open Dataset validate that our RDIoU method can bring substantial improvement for the single-stage 3D object detection.