Exploring YOLO Algorithm application in Smart Traps for Fruit Pest Detection

Agricultural practices are essential for modern civilization and have significantly contributed to population growth in recent years. Food production depends heavily on the natural environment; however, pests can severely damage crops and are often difficult to control. This emphasizes the urgent need for effective pest control measures to protect agriculture. One common way to manage insect pests is by using traps. These traps typically employ bait to attract insects, often consisting of sexual pheromones or food, and they have a mechanism that captures the insect when it attempts to enter. This work explores the training and application of various size variants and versions of the modern object classification algorithm, You Only Look Once (YOLO), during the classification phase of real-time object detection. The focus is on two predominant pests associated with fruit cultivation: Ceratitis capitata and Grapholita molesta. The study utilizes a limited dataset to maximize performance and computational metrics in an intelligent trap-controlled environment.