<![CDATA[Densely populated settlements are one of the urban problems with building density that requires special attention. This research aims to detect and analyze the spatial distribution of building density, especially in detecting building density in residential areas using the Semantic Segmentation deep learning model method with a research dataset sourced from the entire DKI Jakarta Province area. The analysis was conducted using typology criteria in the form of building density levels based on PERMEN PUPR No. 14 of 2018 concerning the Prevention and Improvement of the Quality of Slums and Slum Settlements, which was processed through the Kaggle Notebook and Google Colaboratory platforms using the Python programming language and based on the U-Net architecture. The segmentation results show that using the U-Net architecture is capable of classifying image pixels with an accuracy of 70% in distinguishing between dense and Sparse buildings, which indicates fairly good accuracy performance. The output produced in this final project research is a web interface for detecting dense and Sparse buildings that can be used as a tool to aid in decision-making for regional planning. This research shows that the Semantic Segmentation deep learning model approach can be an efficient and objective solution in satellite image-based spatial analysis. Keywords: Deep Learning, Building Density, Semantic Segmentation ]]>
Copyrights © 2025
