Google Research

Google Research has been at the very forefront of Radiance Field innovation. Of the original NeRF paper authors, Jon Barron, Ben Mildenhall, and Pratul Srinivasan, either were working at Google or went to work at Google following the seminal paper's release. Google Research also now has Georgios Kopanas, one of the original authors of 3D Gaussian Splatting on their team.

On the research side, Google has continued to push the radiance field envelope forward, with important papers such as Mip-NeRF, Mip-NeRF 360, Zip-NeRF, CamP, Nuvo, MERF, SMERF, Dreamfusion, Reconfusion, BakedSDF, Binary Occupancy Grids, Ref-NeRF, CAT3D, NeRF-Casting and more.

CloudNeRF

Google Cloud has recently released CloudNeRF to the public. Like the name implies, it is a cloud based implementation of Google's state of the art method, Zip-NeRF and CamP.

There are currently two implementation of CloudNeRF that people can utilize. There is a JAX framework which runs on JAX for numerical computation and automatic differentiation, prioritizing efficiency and accurate 3D reconstruction from 2D images. Additionally there is a Pytorch implementation too. Google suggests that users utilize the Jax implementation for stronger results.

CloudNeRF is not a free service, but users are able to get $350 in Google Cloud credits by signing up with this link. Note that this is an affiliate link and will greatly help support the sustainability of this website.

Once you have created a Google Cloud account, I have created a tutorial on how to use the platform!

Google has also made a sample dataset available from their Mip-NeRF 360 paper for people to use if they don't have access to custom data.