Computer Science Thesis Proposal

— 11:30am

In Person and Virtual - ET - Reddy Conference Room, Gates Hillman 4405 and Zoom

ARJUN TEH, Ph.D. Student, Computer Science Department, Carnegie Mellon University

Computational Lens Design

From smartphones to microscopes and telescopes, optical lenses can be found in almost every part of modern society. Lens designs themselves span a large gamut of use cases, imaging microscopic phenomena, accommodating wide angle views, and viewing objects astronomically far away from us to name a few. Behind each of these lenses is an expert lens designer who has extensively used design tools that simulate the optics and optimize the designs using brute force methods like finite differencing or random sampling. Designers often need to intervene during the optimization process as these programs fail to produce better designs. 

We present differentiable rendering techniques that efficiently calculates gradients of design objectives with respect to lens design parameters. Specifically, our contribution is twofold. First is a memory efficient method for calculating gradients of optical elements made of heterogeneous materials (so called, GRIN optics). Second is a method for calculating the gradient of light throughput with respect to lens parameters, a quantity that previous methods were not able to optimize directly. In this proposal, we seek to develop methods for searching for high performing designs over non-differentiable parameters in the lens design space, such as number of elements in the device.

Thesis Committee:

Ioannis Gkioulekas (Co-chair)
Matthew O'Toole (Co-chair)
Jim McCann
Bernd Bickel (ETH - Eidgenössische Technische Hochschule Zürich)

Additional Information
In Person and Zoom Participation.  See announcement.

Add event to Google
Add event to iCal