ISSN:1005-3026

A HEURISTIC APPROACHES TOWARDS CITRUS FRUIT AND LEAVES DISEASE DETECTION USING MACHINE LEARNING & ARTIFICIAL INTELLIGENCE TECHNIQUES

Puneet Shetteppanavar1, Dr. Veeranna Kotagi2, Dr. Nazimunisa3, Shaik Rasool Basha4

Computer Science and Engineering, Sree Dattha Institute of Engineering and Science, Hyderabad, INDIA1

Information Science and Engineering, NITTE Meenakshi Institute of Technology, Bangalore, INDIA2

Computer Science and Engineering, Sree Dattha Institute of Engineering and Science, Hyderabad, INDIA3

Computer Science and Engineering, Sree Dattha Institute of Engineering and Science, Hyderabad, INDIA4

punit054@gmail.com1, veerkotagi211@gmail.com2, pnajma22@gmail.com3, sk.rasoolbasha@gmail.com4

Abstract: Citrus fruit and foliage detection using deep learning techniques is an emerging frontier in agricultural technology. This article explores the application of deep learning models for automated recognition and analysis of citrus fruits and leaves in agricultural sector. The primary focus is on utilizing enhanced deep learning Convolutional Neural Networks (CNNs) modelĀ  to accomplish tasks of object detection. The susceptibility of citrus plants to diseases underscores the importance of effective disease management strategies. Various pathogens, such as fungi, bacteria, and viruses, can infect citrus trees and their fruits. These pathogens can cause symptoms ranging from leaf spots and cankers to more severe issues like fruit rot and tree decline. Early identification of ailments in citrus plants aids in thwarting their spread within orchards, thereby reducing financial setbacks for farmers. The dataset contains images of citrus fruits and leaves depicting both healthy specimens and plants afflicted with diseases such as Black spot, Canker, Scab, Greening, and Melanoses. . In this study, we employed deep learning models including CNN, EfficientNet, VGG16, and ResNet. The primary focus was on developing a CNN model tailored to distinguish between healthy citrus fruits and leaves and those affected by prevalent diseases such as black spot, canker, scab, greening, and melanoses. According to the experimental results, the CNN Model excels in a variety of evaluation metrics compared to its competitors. Boasting a training accuracy of 99.66 and test accuracy of 84.40 percent, the CNN Model is an invaluable decision support resource for farmers diagnosing citrus fruit/leaf diseases.

Keywords: CNN, VGG16, ResNet, EfficientNet, Deep Learning, Object Detection