ABSTRACT:
Agriculture remains one of the most critical sectors for sustaining human life, and the health of crops directly influences food production and economic stability. One of the major challenges faced by farmers is the timely and accurate identification of plant diseases, particularly leaf-based infections that can rapidly spread and severely reduce crop yield. Traditional methods of disease detection rely heavily on manual inspection, which is time-consuming, labor-intensive, and often requires expert knowledge. To address these challenges, this project proposes an automated, real-time leaf disease detection and pesticide recommendation system using Raspberry Pi and deep learning techniques.
The developed system integrates a Raspberry Pi with a CSI camera module to continuously capture live images of plant leaves in real-time. These images are processed using a custom-trained YOLOv5 (You Only Look Once version 5) object detection model, which is capable of identifying and classifying different types of leaf diseases with high accuracy and speed. The model has been trained on a dataset containing various diseased and healthy leaf images, enabling it to distinguish between multiple disease classes under different environmental conditions.
Once a disease is detected with a confidence score above a predefined threshold, the system retrieves the corresponding pesticide recommendations from a predefined database. These recommendations are then displayed on an I2C-based LCD module, providing immediate and clear guidance to the user. Simultaneously, detailed information including the detected disease name and recommended pesticides is printed on the terminal for monitoring and logging purposes. The system also visually highlights the detected disease regions on the captured image using bounding boxes and labels, offering intuitive visual feedback.
The software implementation utilizes Python as the primary programming language, along with OpenCV for image processing and frame handling, and PyTorch for running the deep learning model. The system is optimized to run efficiently on the Raspberry Pi platform, ensuring low power consumption and making it suitable for deployment in rural and remote agricultural environments. The use of embedded hardware makes the solution portable, affordable, and easy to operate without requiring advanced technical expertise.
This project significantly reduces the dependency on manual disease diagnosis, minimizes human error, and enables early detection of plant diseases, thereby preventing large-scale crop damage. By providing instant pesticide recommendations, it also assists farmers in taking timely corrective measures, promoting better crop management practices. Furthermore, the system can be extended in the future to support additional crop types, integrate cloud-based data storage, and include mobile or IoT-based monitoring systems.
In conclusion, the proposed leaf disease detection system demonstrates an effective application of artificial intelligence and embedded systems in agriculture, contributing towards precision farming, increased productivity, and sustainable agricultural practices.

OBJECTIVES:
The primary objective of this project is to develop an intelligent and real-time leaf disease detection system using Raspberry Pi and deep learning techniques, which can assist farmers and agricultural users in identifying plant diseases and taking timely corrective actions. The system aims to combine computer vision, embedded systems, and machine learning to create an efficient, cost-effective, and user-friendly solution for precision agriculture.
The specific objectives of the project are as follows:
• To design and develop a real-time leaf disease detection system
The project aims to build a system capable of capturing live images of plant leaves using a camera module and processing them instantly to detect diseases.
• To implement a deep learning-based detection model (YOLOv5)
A custom-trained YOLOv5 model is used to accurately identify and classify different types of leaf diseases with high speed and reliability.
• To integrate the system with Raspberry Pi for edge computing
The objective is to deploy the model on a Raspberry Pi, ensuring that the system operates independently without requiring high-end computational resources or continuous internet connectivity.
• To provide automated pesticide recommendations
Based on the detected disease, the system retrieves suitable pesticide suggestions from a predefined database, helping users take appropriate action.
• To display results on an LCD screen and terminal
The project ensures that the detected disease and corresponding pesticide information are clearly communicated to the user through an I2C LCD display as well as terminal output.
• To achieve a portable and cost-effective solution
The system is designed to be compact, affordable, and easy to deploy in real agricultural environments, especially in rural areas.
• To reduce manual effort and human error in disease identification
By automating the detection process, the system minimizes dependency on manual inspection and improves accuracy.
• To promote precision agriculture and smart farming practices
The project contributes to modern agricultural techniques by enabling data-driven decision-making and efficient crop management.
• To ensure real-time performance and responsiveness
The system is optimized to provide quick detection and immediate feedback, which is crucial for preventing disease spread.