Watch the recording.

Zoom link.

Date: May 29

Time: 2 p.m.

Location: KEC 1001

Abstract: The next generation of versatile, dexterous AI agents must operate in unstructured, real-world environments requiring high-speed 3D perception with millimeter precision. Current 3D imaging techniques struggle to capture scene information when operating in challenging conditions, such as high-speed motion and large illumination changes. In this talk, I will introduce single-photon cameras, an emerging image sensor technology capable of capturing scene information with picosecond temporal resolution and unprecedented dynamic range. While these sensors offer extreme sensitivity, they present a severe data bottleneck, generating hundreds of gigabytes of raw photon data per second. I will discuss our lab’s recent work that bridges this gap through hardware-efficient algorithms optimized for 3D computer vision tasks. We achieve ~100x reduction in data bandwidth without sacrificing reconstruction fidelity in a variety of tasks, such as 3D scene reconstruction and camera localization and mapping.

Bio: Atul Ingle is an assistant professor in the Department of Computer Science at Portland State University. He directs the Portland State Computational Imaging Lab which designs next generation computational cameras and computer vision algorithms, especially in resource-constrained applications. He received the PhD degree in Electrical Engineering from University of Wisconsin-Madison in 2015. After several years in industrial R&D, first at Philips Healthcare (2013, 2014) and then at Fitbit, Inc. (2105–2017), he was a postdoc at UW-Madison from 2017-2021. Broadly, his research interests include computational imaging, computer vision, image sensors, and signal processing. His work on single-photon 3D cameras received the Marr Prize (honorable mention) award at ICCV 2019 and the ICCP Best Paper award in 2023.