Tomato Plant Disease Identification with Artificial Intelligence for Mobile and Edge Devices

Conducted research and developed AI models for real-time plant disease identification on mobile and edge devices, earning distinction in a master's project.

Overview

As part of the requirements for a Master's degree in Data Science at Teesside University, Petgrave.io's data science researcher conducted a comprehensive research project focused on developing an automated plant disease identification solution using artificial intelligence (AI) technology. The goal was to create a unique and practical application that could be deployed on mobile and edge devices to support agricultural communities.

The Challenge

The research project required the data science researcher to:

Conduct a thorough literature review on the use of AI for tomato plant disease identification
Develop a unique hypothesis and design a rigorous validation process, including data collection and analysis
Build, test, and evaluate the effectiveness of AI models, offering recommendations for future improvements

The researcher also had to document the entire research process, develop a thesis, and present the findings to both internal and external supervisors.

Our Approach

Literature Review: The researcher utilised online databases such as ScienceDirect, IEEE, the Teesside University library, and Google Scholar to gather relevant academic literature on the use of AI for plant disease identification. They carefully reviewed references and citations to build a comprehensive understanding of the current state of research in this field.
Model Development and Validation: Using Python libraries like TensorFlow, Scikit-Learn, NumPy, and Pandas, the researcher developed and trained AI models, specifically leveraging the MobileNets architecture for convolutional neural networks. They designed a robust data collection and analysis process to validate the effectiveness of the models.
Documentation and Presentation: The researcher documented the entire research process, including the codebase, which was made available on GitHub. They then developed a comprehensive thesis and prepared a detailed viva presentation to showcase their findings to the internal and external supervisors.

Technologies and Tools

The data science researcher utilised a range of technologies and tools to conduct the research project:

Scientific Writing: LaTeX (using the OpenLeaf template) and BibTeX were used for formatting the thesis and managing citations.
Programming: Python was the primary programming language, with libraries such as TensorFlow, Scikit-Learn, NumPy, and Pandas used for data processing, model development, and evaluation.
AI Frameworks: The researcher leveraged the MobileNets architecture and convolutional neural networks for the automated plant disease identification models.

The Results