Bounding Box Disparity: 3D Metrics for Object Detection With Full Degree of Freedom

TL;DR

This paper introduces Bounding Box Disparity, a comprehensive 3D object detection metric that accounts for all degrees of freedom, with analytic solutions and open-source implementations for improved evaluation accuracy.

Contribution

It derives analytic solutions for 3D bounding box metrics and proposes Bounding Box Disparity as a new continuous evaluation metric for 3D object detection.

Findings

Derived analytic solutions for 3D bounding boxes.

Proposed Bounding Box Disparity as a new evaluation metric.

Provided open-source implementations and extensions.

Abstract

The most popular evaluation metric for object detection in 2D images is Intersection over Union (IoU). Existing implementations of the IoU metric for 3D object detection usually neglect one or more degrees of freedom. In this paper, we first derive the analytic solution for three dimensional bounding boxes. As a second contribution, a closed-form solution of the volume-to-volume distance is derived. Finally, the Bounding Box Disparity is proposed as a combined positive continuous metric. We provide open source implementations of the three metrics as standalone python functions, as well as extensions to the Open3D library and as ROS nodes.