Home

Awesome

NeRFactor

[Paper] [Video] [Project]

teaser

This is the authors' code release for:

NeRFactor: Neural Factorization of Shape and Reflectance Under an Unknown Illumination
Xiuming Zhang, Pratul P. Srinivasan, Boyang Deng, Paul Debevec, William T. Freeman, Jonathan T. Barron
TOG 2021 (Proc. SIGGRAPH Asia)

This is not an officially supported Google product.

Setup

  1. Clone this repository:

    git clone https://github.com/google/nerfactor.git

  2. Install a Conda environment with all dependencies:

    cd nerfactor
conda env create -f environment.yml
conda activate nerfactor

Tips:

Downloads

If you are using our data, metadata, or pre-trained models (new as of 07/17/2022), see the "Downloads" section of the project page.

If you are BYOD'ing (bringing your own data), go to data_gen/ to either render your own synthetic data or process your real captures.

Running the Code

Go to nerfactor/ and follow the instructions there.

Evaluation (New as of 09/10/2022)

We were contacted a few times about the numbers reported in Table 1. Here are the four sequences we used for generating those numbers: drums_3072, ficus_2188, hotdog_2163, lego_3072, all of which have been released (see the Data section above).

For all sequences, we used these validation views: 0,1,2,3,4,5,6,7 and these (uniformly sampled) test views: 49,99,149,199.

Issues or Questions?

If the issue is code-related, please open an issue here.

For questions, please also consider opening an issue as it may benefit future reader. Otherwise, email Xiuming Zhang.

Acknowledgments

This repository builds upon or draws inspirations from this TOG 2015 code release, the NeRF repository, and the pixelNeRF repository. We thank the authors for making their code publicly available.

Changelog