Comparison of Deep Learning Algorithms for Image Segmentation on Satellite Images
Uložené v:
| Názov: | Comparison of Deep Learning Algorithms for Image Segmentation on Satellite Images |
|---|---|
| Autori: | Hüseyin Acemli, Nida Kumbasar |
| Zdroj: | Volume: 12, Issue: 2479-502 Gazi University Journal of Science Part A: Engineering and Innovation |
| Informácie o vydavateľovi: | Gazi University, 2025. |
| Rok vydania: | 2025 |
| Predmety: | Artificial Intelligence (Other), Remote Sensing, Fotogrametri ve Uzaktan Algılama, High-Spatial Resolution, Semantic Segmentation, Deep Learning, Uzaktan Algılama, Yapay Zeka (Diğer), Photogrammetry and Remote Sensing, Computational Modelling and Simulation in Earth Sciences, Yer Bilimlerinde Hesaplamalı Modelleme ve Simülasyon |
| Popis: | Recent advancements in deep learning have significantly contributed to the development of high spatial resolution (HSR) land cover mapping. However, the distinct geographic patterns between urban and rural areas have limited the generalizability of deep learning algorithms across these domains. To address this challenge, separate datasets for rural and urban environments have been proposed in the literature, aiming to achieve more reliable results in real-world applications. In this study, we utilize the publicly available LoveDA HSR dataset for model and parameter comparison. Experiments were conducted on two distinct scenarios: rural and urban areas. The combination of the Adam optimizer, Dice loss function, and UNet++ architecture exhibited the highest performance in both datasets. A weighted average of this combination, based on the number of test samples, was calculated for both groups, yielding a final performance score of 62.14% in terms of mean Intersection over Union (IoU). |
| Druh dokumentu: | Article |
| Popis súboru: | application/pdf |
| ISSN: | 2147-9542 |
| DOI: | 10.54287/gujsa.1664093 |
| Prístupová URL adresa: | https://dergipark.org.tr/tr/pub/gujsa/issue/91864/1664093 |
| Prístupové číslo: | edsair.doi.dedup.....4923376a05f57f09ec7bcffe2c14c9fd |
| Databáza: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstrakt: | Recent advancements in deep learning have significantly contributed to the development of high spatial resolution (HSR) land cover mapping. However, the distinct geographic patterns between urban and rural areas have limited the generalizability of deep learning algorithms across these domains. To address this challenge, separate datasets for rural and urban environments have been proposed in the literature, aiming to achieve more reliable results in real-world applications. In this study, we utilize the publicly available LoveDA HSR dataset for model and parameter comparison. Experiments were conducted on two distinct scenarios: rural and urban areas. The combination of the Adam optimizer, Dice loss function, and UNet++ architecture exhibited the highest performance in both datasets. A weighted average of this combination, based on the number of test samples, was calculated for both groups, yielding a final performance score of 62.14% in terms of mean Intersection over Union (IoU). |
|---|---|
| ISSN: | 21479542 |
| DOI: | 10.54287/gujsa.1664093 |