Roi Ronan
Technion - Israel Institute of Technology
Location:
40 St. George Street, BA5187
Toronto, ON M5S 2E4
Time: 1:00PM – 2:00PM
Title: Robust and fast 3D tomography of natural objects
Abstract: Computed tomography (CT) aims to recover the inner structure of three dimensional volumetric (3D) heterogeneous objects. CT has extensive use in many domains, including medicine, biology, geophysics and fluid dynamics. We seek CT for natural objects, acquired outdoors, in an uncontrolled environment. When imaging clouds, the image formation model involves multiple scattering. This is highly complex, recursive and makes CT non-linear in the unknowns. Moreover, CT of clouds requires multi-view images acquired by satellites, but clouds evolve while the satellites overfly. Hence, we derive spatiotemporal CT of time-varying clouds. Moreover, we accelerate scattering-CT analysis by several orders of magnitude, and make it flexible to perturbed viewing geometries. We achieve this by developing a learning-based network for variable imaging projection cloud tomography (VIP-CT). The ability to do CT in variable geometries is further generalized for in-situ plankton CT. There, images are taken in oceans and lakes, while each specimen appears only once through the system field-of-view, at a random unknown pose and scale. We achieve 3D tomography using a random ensemble of plankton, and also estimate population statistics.
Bio: Roi Ronen is a Ph.D. student at the Technion - Israel Institute of Technology. His research explores the interface between machine learning, physics-based imaging and computational photography. He obtained his B.Sc. in 2019 and a M.Sc. in 2021, both at the Electrical and Computer Engineering Department, Technion, Israel. In 2021 he conducted a summer internship at Amazon Web Service - AI labs. His paper won the JOSA A Emerging Researcher Best Paper Prize for 2021.
Immediately following Roi Ronan’s seminar, there will be another seminar with Sanjeev Koppal from the University of Florida on “Adaptive Attention: Bringing Active Vision into the Camera”.