BME Seminar: Eung-Joo Lee
Monday, October 30, 2023 - 12:00 p.m.
Eung-Joo Lee
Assistant Professor
Electrical and Computer Engineering
Member, BIO5 Institute
University of Arizona
“Deep Learning-Based Image Analysis on Resource-Constrained Systems“
Keating 103
Zoom link | Password: BearDown
Hosts: Dr. Mario Romero-Ortega and Dr. Shang Song
(Instructor permission required for enrolled students to attend via Zoom)
Persons with a disability may request a reasonable accommodation by contacting the Disability Resource Center at 621-3268 (V/TTY).
Abstract:
In recent years, deep learning approaches have achieved high-end performance on various computer vision tasks. Especially, convolutional neural networks (CNNs) are extensively studied and utilized for image analysis. However, deploying deep learning on edge devices to solve practical computer vision tasks involves major challenges. This includes obtaining vast amounts of labeled training data for related downstream tasks, along with careful considerations regarding real-time requirements, memory consumption, and energy budgets for edge-based applications. In this presentation, I primarily focus on two essential factors for implementing deep learning in image analysis: (1) Ensuring the availability of sufficient labeled training data to achieve high accuracy and (2) Designing deep learning systems that leverage computationally efficient models suitable for practical scenarios. For the first topic, I will present data-centric approaches that aim to achieve model robustness through dataset generation. For the second, I will discuss the development of efficient deep neural network models tailored for reliable implementation on edge devices. Throughout my presentation, I will explore the application of embedded computer vision systems and medical imaging in resource-constrained environments.
Bio:
Eung Joo Lee is an assistant professor in the Department of Electrical Engineering at the University of Arizona. Before joining UArizona, he was a postdoctoral research fellow at MGH/Harvard Medical School. He completed his Ph.D. in electrical and computer engineering at the University of Maryland, College Park. During his doctoral studies, he engaged in a research internship at the U.S. Army Research Laboratory. His research aims to develop deep learning systems that utilize computationally efficient models designed specifically for embedded computer vision systems and medical imaging applications. His focus extends to exploring the development of intelligent systems for a diverse range of applications, including unmanned vehicles, face analysis, hyperspectral imaging, and medical imaging application with multimodal learning.