Implementing State-of-the-Art Deep Learning Approaches for Archaeological Object Detection in Remotely-Sensed Data: The Results of Cross-Domain Collaboration

Martin Olivier,Wouter Verschoof-Van Der Vaart

doi:10.5334/jcaa.78

Martin Olivier, Wouter Verschoof-Van Der Vaart

Open Access

https://doi.org/10.5334/jcaa.78

Copy DOI

Abstract

Implementing State-of-the-Art Deep Learning Approaches for Archaeological Object Detection in Remotely-Sensed Data: The Results of Cross-Domain Collaboration

Highlights

Remote sensing has become an essential part of archaeological spatial research, to locate and characterise the surviving physical evidence of past human activity in the landscape (Verhoeven 2017)
Computer Vision and more generally Machine Learning — which in turn falls under the broad category of Artificial Intelligence — has made enormous progress thanks to the advent of Deep Learning techniques, which are based upon Artificial and Convolutional Neural Networks (CNNs; Krizhevsky, Sutskever & Hinton 2012; LeCun, Bengio & Hinton 2015)
1.2 AIM Based on the above, the main aim of this paper is to develop a reliable, accurate, and fast workflow geared towards the detection of multiple classes of archaeological objects in remotely-sensed data, by using the latest developments from other domains, and from Deep Learning object detection in general

Summary

Introduction

Remote sensing has become an essential part of archaeological spatial research, to locate and characterise the surviving physical evidence of past human activity in the landscape (Verhoeven 2017). Computer Vision and more generally Machine Learning — which in turn falls under the broad category of Artificial Intelligence — has made enormous progress thanks to the advent of Deep Learning techniques, which are based upon Artificial and Convolutional Neural Networks (CNNs; Krizhevsky, Sutskever & Hinton 2012; LeCun, Bengio & Hinton 2015). The latter are hierarchically structured algorithms, consisting of multiple layers, which generally comprise a (image) feature extractor and classifier, loosely inspired by the animal visual cortex (Ball, Anderson & Chan 2017). These challenges are not restricted to archaeology but are prevalent in other domains, such as earth observation or environmental sciences (Sumbul et al 2019; Van Etten 2018)

Objectives

Methods

Results

Discussion

Conclusion