NeRF (Neural Radiance Fields)
AI & Emerging Tech
A neural network that learns a 3D scene from photos and can render it from any new viewpoint.
In practice
A neural radiance field learns, from a set of photographs, a function that says what colour and density exist at every point in space — so it can synthesise a view from an angle nobody photographed. It produced startlingly realistic captures and set the direction for the field, though for most practical work Gaussian splatting has since taken over for being far faster to render.
Questions
What is NeRF (Neural Radiance Fields)?
A NeRF is a neural representation of a scene trained from photographs, able to render novel viewpoints with realistic light and depth.
Why do you use NeRF (Neural Radiance Fields)?
It showed that a scene could be learned rather than modelled, which changed what digitising the real world could mean.
How do you use/apply NeRF (Neural Radiance Fields)?
Photograph the scene from many angles, train the network on those views, then query it for any camera position you want to render.
When do you use NeRF (Neural Radiance Fields)?
For research, for the highest-fidelity captures, and as the concept behind newer, faster methods. For shipping, look at splatting first.
Who uses NeRF (Neural Radiance Fields)?
Researchers, technical artists and spatial designers following where 3D capture is going.
