QR Code

Goddard Space Flight Center, Greenbelt, Maryland 20771

ENGINEERING COLLOQUIUM

Wednesday, September 28, 2022, 11:00 AM / Hybrid: On line and Building 3 Auditorium

photo of Dr. Panetta

Dr. Karen Panetta

"The Eye In AI: Human Visual System (HVS) Approaches for Robust Artificial Intelligence"

ABSTRACT -- Artificial Intelligence holds much promise to help us explore, explain and expand our knowledge of the universe. It can be a game changer to advance and expedite applications for autonomous operations to efficiently explore trends in massive data sets. However, there are major challenges we must overcome to fully utilize AI. One challenge is that AI accuracy and reliability is dependent on the quality, diversity and breadth of the ground truth data used to train it. Another challenge is ensuring the results of AI are conducive for human consumption. This includes ensuring that imagery sensor data is optimal for human visual perception, without the need for a human evaluator in the loop. This talk will present Human Visual System (HVS) methods that emulate the way humans evaluate quality. We show how to utilize this concept in computer vision to ensure AI recognition and detection systems produce optimal results for safety, security, and biomedical applications. Quantitative quality metrics based on Human vision and new efficient AI architectures are introduced to address the challenges. Finally, real world applications of these methods will be demonstrated for removing weather anomalies, for detecting disease and show the impact of these methods to increase detection performance for AI recognition systems.

SPEAKER -- Dr. Karen A. Panetta is the Dean for Graduate Education in the Tufts University School of Engineering, a Professor in the Department of Electrical & Computer Engineering with secondary appointments in the Departments of Computer Science and Mechanical Engineering.

While at the Digital Equipment Corporation (DEC), Dr. Panetta she co-invented the first CPU Digital Twin. (A digital twin is like a simulation of a physical system, but more complex. A digital twin may contain data from multiple sensors in the physical system, and it can be used to run multiple types of simulations of the behavior of the physical system.). Her work on digital twins has been adapted for many industrial processes (such as the design of spacecraft), and has won her number of awards from NASA.

Dr. Panetta has also worked in the area of vision and image processing. Her underwater imaging algorithms are used for ocean exploration and for search and rescue. Her parametric logarithmic image processing allows computers to asses image quality as a human observer would, while enhancing imagery for human perception. These methods are used in areas as divergent as agriculture and self-driving vehicles. For Artificial Intelligence (AI) detection and classification systems, her methods ensure optimal image quality data to reduce misclassifications due to invariant image quality without the need for a human evaluator.

Dr. Panetta's high-speed multiplexor design was used to help restore communications in New York City folloqing the 9-11 attacks.

Dr. Panetta holds a BS in computer engineering from Boston University, an MA in electrical engineering from Northeastern University, and a PhD from Northeastern University in electrical engineering, including work on vision and control in robotics.

Dr. Panetta is a Fellow of IEEE, AIAA, and AAAS, among other organization. In 2011, President Obama awarded Karen the nation’s highest award for Engineering, Science and Mathematics Education and Mentoring. Other awards include the Norm Augustine Award from the National Academies of Engineering and Science, American Association of Engineering Societies, and the Women of Vision Award from the Anita Borg Institute. Her outreach activities include hosting the "Nerd Girl Nation" show, an interview series of STEM women role models.




Engineering Colloquium home page: https://ecolloq.gsfc.nasa.gov
Home