Imaging Science
Seminar
Foundations for
Transforming the Imaging
ChAIn
Zoom Link
here
Dr. Derek J. Walvoord
L3Harris, Space and
Airborne Systems Segment Adjunct Professor
College of Science,
Rochester Institute of Technology
Rochester Institute of
Technology
Abstract:
This seminar explores
the profound impact of Machine Learning (ML) and Artificial Intelligence
(AI) on the modern Imaging chAIn while emphasizing consistency in
mathematical foundations. Correlations within the historical progression
of the Imaging field may help indicate its future direction. However,
this seminar attempts to achieve the same goal by instead exploring the
historical progression of correlation itself. From projections, to
transformers, to foundation models and agentic AI for Computer Vision
(CV), consistencies are dissected to ground our understanding.
Additionally, we present a two-pronged approach to improving CV
generalization in open-world settings. First, we use an agentic AI
workflow to simplify and expand the domain over which synthetic data can
be produced. Second, we stress the value of injecting broad domain
knowledge from pretrained vision-language models (VLMs) into solutions
otherwise dominated by raster data processing. These concepts are
reviewed in the context of our research and development in
re-identification (reID) - a task now ubiquitous in multi-camera video
collection settings, such as retail analytics, autonomous vehicles, and
facility security. We demonstrate qualitative gains over component-level
approaches and emphasize foundation model integration into custom CV
applications.
Bio:
Dr. Derek J. Walvoord is
a Fellow and Chief AI Architect for the Space & Airborne Systems
Segment of L3Harris. With over 17 years at the company, Derek has
experience in photogrammetric sensor modeling, rare event analysis, and
Machine Learning for image and video exploitation problems. Derek holds a
BS and PhD in Imaging Science from the Rochester Institute of Technology,
where he has remained actively involved as an adjunct
professor.
Intended
Audience:
Beginners, undergraduates, graduates. Those with
interest in the topic.<
/span>
To request an interpreter, please visit myaccess.rit.edu
END:VEVENT END:VCALENDAR