Date of Award

6-2026

Degree Name

Doctor of Philosophy

Department

Computer Science

First Advisor

Guan Yue Hong, Ph.D.

Second Advisor

Ajay Gupta, Ph.D.

Third Advisor

Hexu Liu, Ph.D.

Keywords

3D image segmentation, computer vision, human-AI collaboration, lightweight deep learning, Mixed Reality (MR) systems, practical artificial intelligence (AI)

Abstract

Deep learning has become the dominant approach for solving vision-centric problems; however, its successful deployment in real-world applications remains limited by high computational cost, data dependency, and insufficient integration with practical and human-centered environments. While state-of-the art deep learning models often achieve impressive performance in controlled settings, they frequently fail to generalize or operate efficiently under deployment constraints such as limited resources, complex data modalities, and real-time interaction requirements. These limitations motivate the need for a deployment-oriented deep learning framework that balances accuracy, efficiency, and practical usability.

This dissertation investigates the design and deployment of efficient deep learning architectures for vision-centric applications, with the central premise that task-aware architectural design and deployment driven optimization can enable deep learning systems to operate reliably across diverse real-world scenarios. Rather than treating individual applications as independent problems, this work develops a coherent methodological framework and evaluates it progressively across multiple vision tasks with increasing deployment complexity.

The dissertation first addresses deep learning–based segmentation and analysis of 3D volumetric images, where high computational cost and data dimensionality pose significant challenges. Efficient encoder–decoder architectures are developed to enable accurate pixel-level segmentation while reducing computational overhead, supporting automated and non-destructive analysis of complex volumetric data.

Building on this foundation, the dissertation then investigates lightweight deep learning architectures for image classification in agricultural applications, focusing on plant disease detection under resource constrained and field-level conditions. Custom convolutional neural network models are designed to achieve competitive classification accuracy with reduced parameter complexity, enabling practical deployment on low-power platforms.

Finally, the dissertation extends the proposed framework to human-centered vision systems through mixed reality integration, addressing the challenges of real-time inference and human–AI interaction in safety-critical environments. A mixed reality–enabled medical vial identification system is developed by embedding an efficient object detection model within a spatial computing environment, demonstrating how deep learning can be effectively integrated into human workflows to improve accuracy, efficiency, and decision-making.

Collectively, this dissertation demonstrates that efficient, deployment-aware deep learning architectures provide a unifying foundation for a broad range of vision-centric applications, from volumetric image analysis to mobile vision systems and mixed reality–assisted human–AI interaction. The contributions of this work advance practical deep learning design principles and offer a transferable framework for developing scalable, robust, and human-centered artificial intelligence systems.

Access Setting

Dissertation-Open Access

Download

Share

COinS