[4eyes] Bako Qualifying Exam

[Steve C. Bako]

Hi everyone,

I have my qualifying exam on

Monday, Sept. 18th @ 9:30 am in HFH Room 4164

All are welcome! Details are below:

Title: Using Deep Learning for Variance Reduction in Monte Carlo Rendering

Abstract:

Monte Carlo path tracers generate the complex and compelling visual effects and animation that we frequently see in movies and television shows. However, it typically takes many hours or even days to produce a single image and having to render many images can significantly delay production workflows and schedules for studios.

The long convergence time for these systems has spurred an abundance of research in all aspects of the rendering pipeline. However, many of the current strategies rely on simple heuristics that have many failure cases and are far from optimal. Moreover, they do not utilize any prior information when rendering a new scene.

In this thesis, we propose to use deep neural networks to leverage knowledge from previous renders and model complex, non-linear relationships, resulting in a robust, data-driven approach that can replace previous heuristics and improve convergence. Specifically, we demonstrate how deep learning can be used for post-rendering denoising and intra-rendering importance sampling and explore avenues for future research in both topics.

Thanks,
Steve

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cs.ucsb.edu/pipermail/ilab-users/attachments/20170915/587e65fc/attachment.html>