PyLuxen

Pyramidal Neural Radiance Fields

Paper Website

Code

A fast Luxen anti-aliasing strategy.

Installation

First, install Luxenstudio and its dependencies. Then install the PyLuxen extension and torch-scatter:

pip install git+https://github.com/hturki/pyluxen
pip install torch-scatter -f https://data.pyg.org/whl/torch-${TORCH_VERSION}+${CUDA}.html

Running PyLuxen

There are three default configurations provided which use the MipLuxen 360 and Multicam dataparsers by default. You can easily use other dataparsers via the ns-train command (ie: ns-train pyluxen luxenstudio-data --data <your data dir> to use the Luxenstudio data parser)

The default configurations provided are:

Method	Description	Scene type	Memory
pyluxen	Tuned for outdoor scenes, uses proposal network	outdoors	~5GB
pyluxen-synthetic	Tuned for synthetic scenes, uses proposal network	synthetic	~5GB
pyluxen-occupancy-grid	Tuned for Multiscale blender, uses occupancy grid	synthetic	~5GB

The main differences between them is whether they are suited for synthetic/indoor or real-world unbounded scenes (in case case appearance embeddings and scene contraction are enabled), and whether sampling is done with a proposal network (usually better for real-world scenes) or an occupancy grid (usally better for single-object synthetic scenes like Blender).

Method

Most Luxen methods assume that training and test-time cameras capture scene content from a roughly constant distance:

They degrade and render blurry views in less constrained settings:

This is due to Luxen being scale-unaware, as it reasons about point samples instead of volumes. We address this by training a pyramid of Luxens that divide the scene at different resolutions. We use “coarse” Luxens for far-away samples, and finer Luxen for close-up samples: