Detection of Chili Plant Pests and Diseases using Yolov5

Abstract

The rapid advancement of technology has led to the development of various innovative techniques that assist humans in numerous domains, including object detection, which serves to identify individual elements within an image. Object detection is widely utilized due to its ability to accurately recognize each component in an image, making it valuable in addressing real-world challenges. One such challenge is the decline in agricultural income resulting from diseases affecting chili plants. The cultivation of chili plants faces several obstacles, including weather-related factors that contribute to the spread of pests and diseases, ultimately reducing chili production. By implementing object detection technology, farmers can easily identify plant diseases through image analysis, enabling timely and effective treatment. This study employs the YOLOv5 algorithm to evaluate the model's performance in detecting diseases in chili plants. The images used were captured using a smartphone camera with a resolution of 3472×3472 pixels. A total of 430 images were utilized, divided into three subsets: training data, validation data, and test data. To obtain the optimal model, the study conducted three experiments using different data distribution ratios: Experiment 1 with a 70:20:10 split, Experiment 2 with a 75:15:10 split, and Experiment 3 with an 80:10:10 split. Among these, the third experiment yielded the best performance, achieving an average test accuracy of 0.947. The corresponding precision, recall, and mean Average Precision (mAP) scores were 0.946, 0.936, and 0.959, respectively