Detection of Postharvest Green Mold (Penicillium digitatum) and Blue Mold (Penicillium italicum) on Citrus Fruit Using Google Teachable Machine

Abstract

The most important aspect of agricultural management is to ensure the yield and health of plants and crops. Citrus plants are frequently growing fruits throughout the world. Citrus diseases directly affect the fruit's quality and reduce its yield. Citrus production is a vulnerable disease to mold and has resulted in substantial economic losses and a reduction in fruit quality. Current methods to detect these diseases, for example, involve time-consuming, expensive, and prone to human error methods like manual field inspection and laboratory analysis. The advancements in artificial intelligence, very specifically in deep learning, have allowed disease detection at greater accuracy and efficiency. Moreover, this study examines the applicability of Google Teachable Machine, an easy-to-use machine learning tool for formulating devices that detect and classify mold diseases in citrus. Training models on images of healthy and infected citrus fruits offers potential for real-time diagnostics for farmers achieving high accuracy (96% average on a curated dataset of approximately 1,500 images, with a held-out test set). The approach provides an economical, practical, and scalable replacement for conventional practice, potentially enabling earlier detection and contributing to reduced crop losses.