Date Approved

5-29-2026

Graduate Degree Type

Thesis

Degree Name

Applied Computer Science (M.S.)

Degree Program

School of Computing and Information Systems

First Advisor

Rahat Ibn Rafiq

Academic Year

2025/2026

Abstract

This project addresses the challenge of image classification for specific plants in indoor farming environments. With a limited dataset, selecting an efficient and cost-effective model becomes crucial.

The traditional approach of using cloud services for model storage and image processing is not only expensive but also inefficient for this use case, especially when considering the continuous flow of data generated by cameras in indoor farming settings. Cloud-based solutions incur high operational costs, which is not ideal for resource-constrained environments. To overcome these limitations and provide a more economical alternative, we propose the use of IoT devices, specifically Raspberry Pi units, to handle the processing locally.

This approach eliminates the need for costly cloud services while still maintaining an effective and scalable system through the use of distillation techniques.

During the period of the thesis, a complete architecture was designed to enable image classification using Raspberry Pi devices. These devices capture images through cameras and send the data to a centralized IoT device for processing. The centralized device performs the classification task, determining whether the plants are healthy or unhealthy, using a continuous learning architecture to ensure accurate results over time. The research and development efforts focused on optimizing the architecture, testing various classification models, and ensuring that the solution is both efficient and practical for real-world deployment in indoor farming environments.

Download

Share

COinS