Lemon disease classification using CNN-based architectures

Abstract

The focus of my study, "Lemon disease classification using CNN-based architecture," is image-based disease detection. Significant progress has been achieved in image processing, and artificial intelligence and its uses are being applied in many design fields. Humanity has already stepped into the digital age. We use sophisticated cameras to take pictures, and the better, more useful, and more productive the outputs and consequences are, the clearer the pictures are. I utilized fresh, totally infected, and infested lemons for this study. In addition to employing neural network (CNN) tools for AI to analyze the findings, I am creating my groupings of features using the VGG16, MobileNetV2, NASNetMobile, ResNet152, EfficientNetV2, and DenseNet201 models. Applications primarily utilizing growth modeling in agricultural production and lemon agronomic research can benefit from the knowledge of lemon features. The direct measuring techniques used in the past were often labor and time-intensive, basic, and not very dependable. Effective techniques for seeing and recognizing exterior illnesses and other aspects provide the foundation of recommended vision. As of right now, image processing algorithms are evolving quickly to identify and recognize certain color frames of afflicted areas in order to diagnose illnesses. Ultimately, the afflicted region is isolated from the picture. The diseases that impact fruit production, such as citrus canker disease, citrus rust mite disease, and delicious lemons, were then recognized from pictures of lemons. The results of the VGG16, MobileNetV2, NASNetMobile, ResNet152, EfficientNetV2, and DenseNet201 models in my example demonstrate that this autonomous vision-based system performs admirably, with the highest results in VGG16 coming in at 91%.