A pytorch implementation of "Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields"

Paper Website



First, install nerfstudio and its dependencies. Then run:

pip install git+
pip install git+

Finally, install torch_scatter corresponding to your cuda version(

Running Model#

ns-train zipnerf --data {DATA_DIR/SCENE}


Zipnerf combines mip-NeRF 360’s overall framework with iNGP’s featurization approach. Following mip-NeRF, zipnerf assume each pixel corresponds to a cone. Given an interval along the ray, it construct a set of multisamples that approximate the shape of that conical frustum. Also, it present an alternative loss that, unlike mip-NeRF 360’s interlevel loss, is continuous and smooth with respect to distance along the ray to prevent z-aliasing.