Denoising Renderings using Deep Learning


This project builds a Monte Carlo path tracing renderer with AI denoising. Monte Carlo path tracing is a powerful method to render Computer Generated Imagery (CGI), but it is resource intensive and produces excessive noise in images, therefore we propose to remove noise using deep learning. We aim to develop a rendering engine and a denoising engine, and integrate them into an graphics user interface that can load scenes and render them in real-time, focusing on efficiency in computational resources and minimal dependency. Developers will comprehend text like academic papers and e-books, develop the rendering and denoising engines, and build the GUI scene viewer.

We are looking for developers who demonstrate strong passion and willingness to collaborate throughout the yearlong project and have experience in PyTorch and/or C/C++. Experience/interest with 3D computer graphics and fairly complex CNN models is a plus. Email or DM spirulae in the UTMIST Discord if you are interested.