Classification of beet plant diseases using deep learning

Sugar has an essential importance for human life. As a staple food, sugar beet has great agricultural and economic importance with its other by-products besides sugar production. The main reasons that negatively affect the yield and quality of the sugar to be produced are the diseases that may occur in the plant. Due to leaf diseases in sugar beet, root development cannot reach the desired levels, which reduces the sugar content and yield of the plant. Detecting diseases in the leaves at an early stage minimizes the spread of the disease or the damage to the plant. In the detection of diseases, deep learning methods are used for image and object detection along with advancing technology. Among these models, YOLO (You Only Look Once) attracts attention with its constantly updated versions and increasing usage areas. In this study, the classification of Cercospora (leaf spot) and powdery mildew diseases, which are common sugar beet diseases in our country, were examined by using YOLOv8, ResNet50, and DenseNet121 models. With the classification processes performed with YOLOv8 submodels, the training results of ResNet50 and DenseNet121 transfer learning models were compared according to performance evaluation criteria.