Real-Time Multiclass Detection of Citrus Leaf Diseases Using an Enhanced YOLOv11 Architecture

Project Code :TCMAPY2438

Objective

The objective of this project is to develop a real-time citrus leaf disease detection system using deep learning techniques, specifically leveraging the YOLOv12 architecture. This system is designed to address the challenges faced by citrus farming in Pakistan, particularly in Punjab, where citrus crops suffer from diseases like citrus canker, greasy spot, and leaf miner. The aim is to build an automated system that can accurately detect these diseases in real time, significantly improving the efficiency and accuracy of disease detection compared to traditional manual methods. By utilizing a deep learning-based approach, the system will not only reduce human intervention but also provide farmers with immediate feedback, allowing them to take timely and informed actions to prevent further spread of the diseases. The proposed system will process images of citrus leaves in real-time and deliver high-precision disease classifications, providing valuable support for farmers in managing their orchards more effectively. Ultimately, the goal is to contribute to enhancing citrus production, reducing losses, and supporting sustainable agricultural practices in the region.

Abstract

In light of the global challenges surrounding food security, rapid and accurate disease detection is critical for safeguarding agricultural productivity. Citrus, a vital crop in the Pakistani economy, faces declining yields and quality due to foliar diseases. Traditional methods of disease detection rely heavily on manual labor and subjective assessment, which can be time-consuming and prone to errors. This paper proposes a real-time Citrus Multiple Leaf Disease Detection (CLDD) system, leveraging the advanced YOLOv12 architecture. The dataset used for training the model consists of 866 images, including citrus canker, greasy spot, and leaf miner diseases collected from Okara orchards in Punjab, Pakistan. Data augmentation was employed to enhance the dataset's generalization capability. The YOLOv12 model demonstrated exceptional performance, achieving a precision of 99.2%, recall of 99.5%, and a mean Average Precision (mAP50) of 99.2%. The system processes images in real-time with a processing time of just 19.7 ms per image. Ablation studies revealed that incorporating a CSPDarknet53 backbone, a PANet neck, and advanced augmentation techniques significantly improved the model’s robustness. The results demonstrate the potential of deep learning models in automating real-time citrus disease detection, providing an accurate and scalable solution to address the challenges faced by the agricultural sector in Pakistan and beyond.

Keywords: Citrus Multiple Leaf Disease Detection (CMLDD), Deep Learning, Machine Learning, YOLOv12

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.