From concrete jungles to urban gardens: AI-powered solutions for sustainable food production in cities

Abstract

Introduction: Urban agriculture in Indonesia faces critical challenges including agricultural land conversion, aging farmer workforce (39% over 55 years, only 21% millennials), and rural urban inequality. While deep learning technologies prove effective for agricultural optimization, Indonesia lags neighboring countries due to regulatory ambiguity, limited incentives, and low youth participation. This study develops Urfalogy, an artificial intelligence powered platform addressing three primary urban farming constraints: limited space, insufficient capital, and inadequate technology. Methods: This research employed Agile software development methodology integrated with deep learning. The You Only Look Once version 8 (YOLOv8) algorithm was utilized for environmental object detection and segmentation. Dataset preprocessing included multiple augmentation techniques: scaling, geometric transformation, brightness adjustment, contrast and color saturation modifications. The platform integrates nine features: artificial intelligence layout designer, plant variety recommender, plant health detection, soil monitoring with internet of things sensors, e-commerce, real time expert consultation, appointment scheduling, interactive tutorials, and analytics dashboard. Finding: Model training achieved optimal performance metrics at epoch 100: segment loss of 0.56756, recall of 90.01%, and mean Average Precision at intersection over union 0.50 (mAP50) of 90.715%. During inference, the model successfully identified environmental components (ceiling, wall, floor), enabling precise spatial mapping for garden layout design. The integrated platform demonstrates comprehensive end to end capability supporting complete urban farming workflow from planning through sales. Conclusion: Urfalogy represents a transformative solution effectively bridging Indonesia's urban agriculture gap through artificial intelligence, Internet of Things integration, and human centered design, significantly advancing sustainability, food security, and economic opportunities. Novelty/Originality of this article: This research uniquely combines deep learning-based spatial optimization with comprehensive platform ecosystem design, integrating YOLOv8 environmental analysis with real-time consultation and e-commerce, addressing specific technological, economic, and accessibility barriers in Indonesian urban agriculture.