Centerpoints Are All You Need in Overhead Imagery

TL;DR

This paper introduces centerpoint-based architectures for overhead object detection, demonstrating that they perform nearly as well as more detailed labeling methods, thus reducing labeling effort.

Contribution

It proposes novel single- and two-stage network architectures utilizing centerpoints, showing they match the performance of detailed labeling approaches.

Findings

Centerpoint architectures achieve comparable accuracy to detailed labeling methods.

Using centerpoints reduces labeling complexity and effort.

Performance is validated on three overhead object detection datasets.

Abstract

Labeling data to use for training object detectors is expensive and time consuming. Publicly available overhead datasets for object detection are labeled with image-aligned bounding boxes, object-aligned bounding boxes, or object masks, but it is not clear whether such detailed labeling is necessary. To test the idea, we developed novel single- and two-stage network architectures that use centerpoints for labeling. In this paper we show that these architectures achieve nearly equivalent performance to approaches using more detailed labeling on three overhead object detection datasets.