TY - JOUR

T1 - Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs

AU - Kazimi, Bashir

AU - Sester, Monika

N1 - Funding Information: This research was funded by the Lower Saxony Ministry of Science and Culture through the “Niedersächsisches Vorab” funding initiative and with the collaboration of Lower Saxony State Office for Heritage (Niedersächsisches Landesamt für Denkmalpflege). The publication of this article was funded by the Open Access Fund of the Leibniz Universität Hannover.

PY - 2023

Y1 - 2023

N2 - Deep learning models need a lot of labeled data to work well. In this study, we use a Self-Supervised Learning (SSL) method for semantic segmentation of archaeological monuments in Digital Terrain Models (DTMs). This method first uses unlabeled data to pretrain a model (pretext task), and then fine-tunes it with a small labeled dataset (downstream task). We use unlabeled DTMs and Relief Visualizations (RVs) to train an encoder-decoder and a Generative Adversarial Network (GAN) in the pretext task and an annotated DTM dataset to fine-tune a semantic segmentation model in the downstream task. Experiments indicate that this approach produces better results than training from scratch or using models pretrained on image data like ImageNet. The code and pretrained weights for the encoder-decoder and the GAN models are made available on Github.1

AB - Deep learning models need a lot of labeled data to work well. In this study, we use a Self-Supervised Learning (SSL) method for semantic segmentation of archaeological monuments in Digital Terrain Models (DTMs). This method first uses unlabeled data to pretrain a model (pretext task), and then fine-tunes it with a small labeled dataset (downstream task). We use unlabeled DTMs and Relief Visualizations (RVs) to train an encoder-decoder and a Generative Adversarial Network (GAN) in the pretext task and an annotated DTM dataset to fine-tune a semantic segmentation model in the downstream task. Experiments indicate that this approach produces better results than training from scratch or using models pretrained on image data like ImageNet. The code and pretrained weights for the encoder-decoder and the GAN models are made available on Github.1

KW - Archaeology

KW - Convolutional Neural Networks

KW - Deep Learning

KW - Digital Terrain Models

KW - Generative Adversarial Networks

KW - Relief Visualization

KW - Self-Supervised Learning

UR - http://www.scopus.com/inward/record.url?scp=85179966569&partnerID=8YFLogxK

U2 - 10.5334/jcaa.110

DO - 10.5334/jcaa.110

M3 - Article

AN - SCOPUS:85179966569

VL - 6

SP - 155

EP - 173

JO - Journal of Computer Applications in Archaeology

JF - Journal of Computer Applications in Archaeology

IS - 1

ER -