Computer Vision-Based Identification and Classification of Leaf Diseases in Longyard Beans

Emerging plant infections are a considerable problem for global farming, resulting in the need for improved disease detection methods. The study is about leaf diseases of Longyard Beans as addressed using advanced deep learning models – VGG-16, VGG-19, ResNet-50. The proposed work seeks to improve the accuracy and reliability of automatic disease identification in the agricultural domain by utilizing these models for leaf disease identification in longyard beans. A rich dataset composed of pre-processed leaf images from Longyard Beans is used to facilitate effective model training. The images undergo augmentation to increase the sample size followed by image enhancement to remove noises in the image. Further, the images are annotated and the models are used to identify the bean leaves. Comparative analysis between VGG-16, VGG-19, and ResNet-50 illustrates their relative strengths and drawbacks. The models are evaluated using a confusion matrix and ROC/AUC curve. The accuracies obtained are 94 %, 92 %, and 91 % for VGG-16, VGG-19, and ResNet-50 respectively.