Learning and Scoring Point Process Models for Object Detection in Satellite Images

TL;DR: This paper proposes a joint Point Process and CNN method for object detection in satellite images, leveraging interaction priors to complement limited visual information, with matching parameter estimation and result scoring procedures for improved accuracy and interpretability.

Abstract: In this paper we propose a joint Point Process and CNN based method for object detection in satellite imagery. The Point Process allows building a lightweight interaction model, while the CNN allows to efficiently extract meaningful information from the image in a context where interaction priors can complement the limited visual information. More specifically, we present matching parameter estimation and result scoring procedures, that allow to take into account object interaction. The method provides good results on benchmark data, along with a degree of interpretability of the output. The code will be available at github.com/Ayana-Inria/