Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs
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-...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Ubiquity Press
2023-11-01
|
| Series: | Journal of Computer Applications in Archaeology |
| Subjects: | |
| Online Access: | https://account.journal.caa-international.org/index.php/up-j-jcaa/article/view/110 |
| _version_ | 1827576374397763584 |
|---|---|
| author | Bashir Kazimi Monika Sester |
| author_facet | Bashir Kazimi Monika Sester |
| author_sort | Bashir Kazimi |
| collection | DOAJ |
| description | 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 |
| first_indexed | 2024-03-08T21:10:16Z |
| format | Article |
| id | doaj.art-4935ba563f7d443e9e0d9137499809d3 |
| institution | Directory Open Access Journal |
| issn | 2514-8362 |
| language | English |
| last_indexed | 2024-03-08T21:10:16Z |
| publishDate | 2023-11-01 |
| publisher | Ubiquity Press |
| record_format | Article |
| series | Journal of Computer Applications in Archaeology |
| spelling | doaj.art-4935ba563f7d443e9e0d9137499809d32023-12-22T06:35:31ZengUbiquity PressJournal of Computer Applications in Archaeology2514-83622023-11-0161155–173155–17310.5334/jcaa.11032Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMsBashir Kazimi0https://orcid.org/0000-0003-1802-7511Monika Sester1https://orcid.org/0000-0002-6656-8809Leibniz University Hannover Institute of Cartography and Geoinformatics Appelstr. 9a, 30167 HannoverLeibniz University Hannover Institute of Cartography and Geoinformatics Appelstr. 9a, 30167 HannoverDeep 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.1https://account.journal.caa-international.org/index.php/up-j-jcaa/article/view/110self-supervised learningdigital terrain modelsdeep learningarchaeologyconvolutional neural networksgenerative adversarial networksrelief visualization |
| spellingShingle | Bashir Kazimi Monika Sester Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs Journal of Computer Applications in Archaeology self-supervised learning digital terrain models deep learning archaeology convolutional neural networks generative adversarial networks relief visualization |
| title | Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs |
| title_full | Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs |
| title_fullStr | Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs |
| title_full_unstemmed | Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs |
| title_short | Self-Supervised Learning for Semantic Segmentation of Archaeological Monuments in DTMs |
| title_sort | self supervised learning for semantic segmentation of archaeological monuments in dtms |
| topic | self-supervised learning digital terrain models deep learning archaeology convolutional neural networks generative adversarial networks relief visualization |
| url | https://account.journal.caa-international.org/index.php/up-j-jcaa/article/view/110 |
| work_keys_str_mv | AT bashirkazimi selfsupervisedlearningforsemanticsegmentationofarchaeologicalmonumentsindtms AT monikasester selfsupervisedlearningforsemanticsegmentationofarchaeologicalmonumentsindtms |