Optimizing Urban Farming Potential Using Google Street View and Machine Learning

Google Street View (GSV) and machine learning offer powerful tools for optimizing urban farming potential by providing valuable insights into site suitability. This study integrates GSV imagery with machine learning techniques, specifically Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs), to assess urban farming suitability in Bandung City. Using ResNet models for building and road classification, we achieved an initial accuracy of 0.88, which improved to 0.97 with SVM optimization, demonstrating the model’s effectiveness in identifying optimal urban farming locations. Our framework estimates potential urban farming areas in Bandung, revealing a range from 21,882 m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to 48,784 m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (2.19 ha to 4.88 ha), with the highest potential in suburban districts such as Lengkong and Astanaanyar. The study highlights the utility of machine learning for urban planning applications, with broader implications for sustainable city development. Future work will expand the dataset for greater generalizability and explore additional applications of GSV imagery to enhance urban management insights.