Potato Leaf Disease Detection Using Faster R-CNN and YOLO Models

Authors : Sara Medojević

Pages : 144-150

View : 29 | Download : 23

Publication Date : 2024-12-22

Article Type : Research Paper

Abstract :Potato is one of the most important food crops globally in terms of total food production, significantly impacting the global economy. Infected potato plants show visible symptoms on their leaves, which drastically simplifies the process of early detection, disease prevention, and minimizing the risk to uninfected plants. Smart farming and new advanced technologies incorporate different tools for real-time monitoring and analysis. Most of the models used for potato leaf disease detection are based on Deep Learning architectures, most commonly on Convolutional Neural Network (CNN) architecture, which is suitable for computer vision and image recognition. This paper depicts and compares the performances of the YOLOv11 Object Detection (Fast) model, YOLOv11s model, and Faster R-CNN X101-FPN model. These models were trained on a dataset developed for object detection in Roboflow. This dataset consists of 1200 images and 1500 annotations. A single object was labeled as one of the six classes: Pest, Bacteria, Fungi, Healthy, Phytophthora, and Nematode. Performance metrics show that these models achieve reputable results without excessive training time, making them suitable for real-time monitoring systems. YOLOv11 Object Detection (Fast), YOLOv11s, and Faster R-CNN X101-FPN achieved mAP50 scores of 95.1%, 97.6%, and 92.62%, respectively.
Keywords : YOLO modelleri, Faster R-CNN modeli, Roboflow, nesne tespiti, patates yaprağı hastalığı