A YOLOV8 CLASSIFICATION APPROACH USING ESPCN PROCESSED IMAGES FOR BEAN LEAF DISEASE DETECTION

Abstract

Agriculture is crucial for economic growth, with automation emerging as a global concern due to the increasing population and food demand. Beyond simple cultivation, agriculture is vital to communities, the foundation of economies, and the first line of defence against the threat of food insecurity. Plant diseases pose a constant threat to this vital industry, while they also present several other difficulties. Plant diseases also pose a threat to crop productivity and the fragile balance that exists between food production and the growing needs of an expanding world population. Beans are a vital crop that has served mankind as a nutritious crop throughout history. This research work has classified bean leaf diseases of angular leaf spot, bean rust, and healthy bean leaves using Artificial Intelligence fused technologies. While testing on preprocessing methods, image-enhancing algorithms such as Fast super resolution Convolutional Neural Networks (FSRCNN), Efficient Sub-Pixel Convolutional Neural Networks (ESPCN), and Enhanced Super-Resolution Generative Adversarial Networks (ESRGAN) were used to boost image quality in the respective training phases. Out of the pre-trained Convolutional Neural Networks (CNN) such as MobileNet, MobileNetV2, EfficientNetB0 - B6 models, NasNet, and You Only Look Once - Version 8 (YOLOv8) model, the YOLOv8 model gave an impressive model accuracy of 99.6% when the images were enhanced using ESPCN algorithm while classifying the images accurately. YOLOv8 proved the best model performance by scoring optimal metric values compared to CNN models. Therefore, it can be concluded that YOLOv8 with ESPCN can be used to efficiently classify bean leaf diseases and achieve promising results.