Speaker: Nathan Jessurun (under Navid Asadi)
Abstract: Printed Circuit Boards (PCBs) can contain anywhere from hundreds to thousands of surface-mount devices (SMDs) and silkscreen markings. Such information is readily visible in optical images of the board’s surface and is useful for various forms of hardware assurance such as Bill-of-Material generation, inspection, and verification. However, extracting relevant components and marking information is a challenging task due to the large image sizes, complex component shapes, and significant number of components present. This training introduces S3A: a software application designed to expedite labeling regions of interest in these cases. Discussion will cover the trade-offs between label quality and quantity as it relates to PCB annotation, factors to consider during data clustering/classification, and the benefits of integrating a semi-automated annotation workflow.
Short bio: Nathan Jessurun graduated from Cedarville University in 2019 with a B.S. degree in Computer Engineering. Shortly after, he enrolled in the PhD program at the University of Florida’s FICS Research Lab under Dr. Navid Asadi. His current research interests involve object localization and characterization in multiple image modalities, from THz to optical to X-ray wavelengths. He has several active projects that use image segmentation techniques for hardware assurance purposes such as Bill-of-Materials generation and device fingerprinting.
Bookings are closed for this event.