A Streamable Dynamic Scene Representation with Decomposed Neural Radiance Fields

Paper Website

Nerfstudio add-on code

NeRFPlayer Video


First install nerfstudio dependencies. Then run:

pip install git+

Running NeRFPlayer#

Details for running NeRFPlayer can be found here. Once installed, run:

ns-train nerfplayer-ngp --help

Two variants of NeRFPlayer are provided:




NeRFPlayer with nerfacto backbone


NeRFPlayer with instant-ngp-bounded backbone

Method Overview#

method overview
First, we propose to decompose the 4D spatiotemporal space according to temporal characteristics. Points in the 4D space are associated with probabilities of belonging to three categories: static, deforming, and new areas. Each area is represented and regularized by a separate neural field. Second, we propose a hybrid representations based feature streaming scheme for efficiently modeling the neural fields.

Please see TODO lists to see the unimplemented components in the nerfstudio based version.