Abstract

Neural Radiance Fields (NeRFs), presented in ECCV 2020 just two years ago demonstrated exciting potential for photo-realistic and immersive 3D scene reconstruction from a set of calibrated images. It was followed by a surge of works that explore the potential of using Neural Volumetric Rendering as a technique for enabling many exciting applications and addressing fundamental problems in Computer Vision, Graphics, Robotics and more.
In this tutorial, we will present a short course that approaches Neural Volumetric Rendering from the first principles, including its relation to the history of image based rendering and inverse graphics, its core components and their derivations, common practices, future challenges, and hands-on coding examples. The goal of this half-day tutorial is not to present a series of talks on recent papers in this area, but to provide pedagogical building blocks for novice and intermediate researchers to deeply understand the material by abstracting away the recent developments in the area of Neural Volumetric Rendering.

Organizers