An Evaluation of Applied ML Techniques in Mold Detection and Prediction in Mung Bean Sprouts
Conference
10th International Conference on Agricultural Statistics
Format: CPS Abstract - ICAS 2026
Abstract
The mung bean sprout industry is crucial for many small businesses in the Philippines yet frequently hindered by molding caused by contamination and inadequate quality control practices. Existing methods such as traditional experience-based methods and resource-intensive sterilization, either prove unreliable or compromise the sprouts' physical and nutritional quality. To combat this, a machine learning-based system was developed and evaluated, integrating a hardware setup with a Raspberry Pi 5, a DHT11 temperature and humidity sensor, and a webcam to non-invasively monitor sprout cultivation. The research employed two primary algorithms: the Isolation Forest algorithm for predicting mold growth based on anomalous environmental data and a custom You Only Look Once v11 nano (YOLOv11n) model for image-based mold detection to validate and confirm mold growth visually. The findings demonstrate that this data-driven approach effectively identifies mold presence and pinpoints environmental conditions that signal a high risk of mold development. The application these techniques offer a scalable solution for automated quality assessment, substantially reducing economic losses for microentrepreneurs, enhance public health, and support the development of intelligent agricultural systems.