2022 Thayer Investiture

All Thayer Events

Tackling Real-World Constraints with End-to-End Computational Vision Systems

Feb

08

Feb 08

Zoom - contact Susan Cable

Abstract:
Rapidly growing machine learning techniques such as deep learning have produced powerful computer vision algorithms. However, these algorithms usually apply to images and videos captured with traditional camera designs that have been principally unchanged for decades. Furthermore, real-world applications such as robotics, autonomous navigation, augmented/virtual reality, IoT, and healthcare need systems that adhere to fundamental constraints such as size, weight, power, and privacy. These fundamental constraints cannot be addressed by a software-only solution but demand a joint hardware-software solution. In my talk, I will present end-to-end computational vision systems that execute "computation" at all stages of a physical vision system, from optics to sensors to algorithms. By employing a holistic approach to design, I will showcase various physical vision systems that tackle the constraints of size, weight, power, and privacy. These systems are built with custom optics and sensor chips, manufactured by accessible nanofabrication techniques, and machine learning methods to design and extract information. Finally, I will discuss how the holistic approach of end-to-end computational vision systems can impact potential future applications.

Bio:
Dr. Vivek Boominathan is a postdoctoral research associate in the Department of Electrical and Computer Engineering at Rice University. He received his Ph.D. in 2019, advised by Prof. Ashok Veeraraghavan, and co-advised by Prof. Jacob Robinson and Prof. Richard Baraniuk. His research interests lie at the intersection of computer vision, machine learning, applied optics, and nanofabrication. His contributions have appeared in a broad spectrum of venues such as Science Advances, Nature BME, IEEE journals, optics journals, vision conferences, and circuits conferences. He has also published a review article, in Optica, around his Ph.D. topic of lensless imaging. His work has been covered by news media such as EurekAlert, NPR, Phys.org, and NDTV India. He has co-organized a tutorial on Computational Imaging and Machine Learning in CVPR 2019 and serves as the publication co-chair for International Conference on Computational Photography (ICCP) since 2020. More details of Dr. Boominathan can be found at https://vivekboominathan.com/.

Contact

For more information, contact Susan Cable at susan.perry.cable@dartmouth.edu.