MKIoU Loss: Towards Accurate Oriented Object Detection in Aerial Images

TL;DR

This paper introduces MKIoU, a novel loss function for oriented object detection in aerial images, addressing boundary issues and improving alignment with evaluation metrics through Gaussian modeling and modulation techniques.

Contribution

The paper proposes MKIoU, a modulated Kalman IoU loss with Gaussian modeling and a new Gaussian Angle Loss to enhance oriented bounding box regression accuracy.

Findings

Improved detection accuracy on DOTA, UCAS-AOD, and HRSC2016 datasets.

Effective handling of boundary problems and angle confusion in oriented bounding boxes.

Enhanced consistency between loss function and evaluation metrics.

Abstract

Oriented bounding box regression is crucial for oriented object detection. However, regression-based methods often suffer from boundary problems and the inconsistency between loss and evaluation metrics. In this paper, a modulated Kalman IoU loss of approximate SkewIoU is proposed, named MKIoU. To avoid boundary problems, we convert the oriented bounding box to Gaussian distribution, then use the Kalman filter to approximate the intersection area. However, there exists significant difference between the calculated and actual intersection areas. Thus, we propose a modulation factor to adjust the sensitivity of angle deviation and width-height offset to loss variation, making the loss more consistent with the evaluation metric. Furthermore, the Gaussian modeling method avoids the boundary problem but causes the angle confusion of square objects simultaneously. Thus, the Gaussian Angle Loss (GA Loss) is presented to solve this problem by adding a corrected loss for square targets. The proposed GA Loss can be easily extended to other Gaussian-based methods. Experiments on three publicly available aerial image datasets, DOTA, UCAS-AOD, and HRSC2016, show the effectiveness of the proposed method.