Virtual reality aims to immerse users in believable simulated environments, but creating sufficiently detailed and realistic VR worlds is challenging. An emerging technique called neural radiance fields could revolutionize the level of realism possible in VR spaces.
What are Neural Radiance Fields?
Neural radiance fields (NeRFs) are a method of digitally representing a 3D scene using neural networks. Developed by researchers at UC Berkeley and Google Research, NeRFs can synthesise highly realistic novel views of real world scenes.
NeRFs work by assigning emitted color values called radiance to points throughout the space of the scene. These values are predicted by a neural network trained on images of the scene.
The neural network effectively encodes the scene’s geometry, lighting, textures, and more as weighted values between inputs and outputs. With the scene mapped as coloured points cloud, NeRFs can render photorealistic images from any viewpoint.
Benefits for VR Environments
NeRFs have huge implications for crafting virtual worlds. Their ability to convincingly reconstruct 3D environments from images makes modelling realistic VR spaces or people far more accessible.
NeRF models also require less storage space compared to traditional 3D graphics data.
VR developers can scan a real location, train a NeRF on the images, and reconstruct it with great precision in VR. This works for both outdoor cityscapes and indoor rooms. NeRFs even allow editing the scene by altering the neural network’s weights.
While still an emerging technique, NeRFs demonstrate the growing power of neural networks to model the physical world for VR. As the technology improves, NeRFs could enable VR users to explore digitized versions of real spaces with new heights of realism and immersion.
The future of NeRFs?
Did anyone say NeRFs as a service (NaaS!).
Apple and Google with its already in-play network of small smart cars and cameras mapping our roads for navigation could, and will be major players in this technology.
Money >Infrastructure > Audience, they could be unstoppable in this field.