Study of Deep Learning Models for Automatic Lemon Plant Disease Detection and Classification

Abstract

Agriculture is the backbone of Indian economy. Early detection of disease in plant is very newlineimportant for preventing the economic losses and increasing the productivity of the fruits. newlineThis is the one of the reasons that disease detection in lemon plants plays an important newlinerole in agriculture field. There are three types of disease fungal bacterial and viral present newlinein lemon, like Black spot, Canker, Melanose, Greening, scab etc. Lemon fruit diseases newlineare a major threat to food security, which will lead to productivity loss, economic loss, newlinequality loss and quantity loss. There are minimal numbers of technologies that have been newlinedeveloped to assist farmers throughout the world, but if automated detection techniques newlineare employed, it will require less effort, take less time, and be more accurate. The newlinesymptoms can be seen on plant parts such as the leaves, stems, and fruits. Predicting newlinedisease stage is challenging in many regions of the world due to a lack of technologies. newlineDue to the shape, colour, and texture of different species, lemon disease detection and newlineclassification remain challenging. Detection of plant disease through some automatic newlinemethod is useful as it reduces a huge work of monitoring in large farms of crops, and at newlinevery early phase itself, it detects the symptoms of diseases i.e., when they appear on plant newlineleaves. This research presents different deep learning (DL) technique for lemon plant newlinedisease detection to achieve a great potential in terms of increasing accuracy. newline newlineDiseases have a significant impact on lemon fruit production. This is one of the factors newlinethat contributed to the significance of plant disease detection in the agriculture. Machine newlinelearning (ML) models use a variety of techniques for classification and detection, but newlinedeep learning (DL) may be the most accurate due to the smartness of computer vision newlineresearch. The AlexNet, GoogleNet, ResNet and SqueezeNet models with and without newlinedata augmentation are employed in this study. Data augmentation involves the technique newlineof generating new data poin