Gaussian Bounding Boxes and Probabilistic Intersection-over-Union for Object Detection

TL;DR

This paper introduces a probabilistic approach to object detection using Gaussian distributions to represent object regions, along with a new similarity measure called ProbIoU based on Hellinger Distance, improving alignment with segmentation masks.

Contribution

It proposes a fuzzy Gaussian-based object representation and a ProbIoU similarity measure, enabling better object localization and seamless integration into existing detectors.

Findings

Gaussian representations are closer to segmentation masks

ProbIoU-based loss functions improve detection accuracy

Seamless mapping from traditional bounding boxes to Gaussian models

Abstract

Most object detection methods use bounding boxes to encode and represent the object shape and location. In this work, we explore a fuzzy representation of object regions using Gaussian distributions, which provides an implicit binary representation as (potentially rotated) ellipses. We also present a similarity measure for the Gaussian distributions based on the Hellinger Distance, which can be viewed as a Probabilistic Intersection-over-Union (ProbIoU). Our experimental results show that the proposed Gaussian representations are closer to annotated segmentation masks in publicly available datasets, and that loss functions based on ProbIoU can be successfully used to regress the parameters of the Gaussian representation. Furthermore, we present a simple mapping scheme from traditional (or rotated) bounding boxes to Gaussian representations, allowing the proposed ProbIoU-based losses to be seamlessly integrated into any object detector.