Introduction Page
What is a NeRF?
Neural Radiance Fields (NeRFs) are relatively new 3D modeling frameworks which can learn fully continuous 3D representations of scenes captured from a handful of 2D images. More specifically, these models train a neural network to describe the radiance field present within a scene. Originally proposed by Mildenhall et al. (2020), NeRFs have come a long way in just about every aspect of their performance (speed, size, accuracy, etc.), and are in the perfect spot for researchers to mature them into powerful tools for use in the aerospace industry.
This webpage will give a detailed description of how NeRFs work, starting with the original 2020 model, and the models I have changed/created so far. I will try my best to outline the most important aspects of these models I have learned in order to prevent most of the growing pains I encountered initially. I recommend covering the section describing how these models are made first, but if you are already familiar you may go ahead and check out my work and how to use it.
Fun teaser image:
