Persistent Nature

Project Page | Paper | Bibtex

Persistent Nature: A Generative Model of Unbounded 3D Worlds.
CVPR 2023
Lucy Chai, Richard Tucker, Zhengqi Li, Phillip Isola, Noah Snavely

Please note that this is not an officially supported Google product.

Prerequisites

• Linux
• gcc-7
• Python 3
• NVIDIA GPU + CUDA CuDNN

Table of Contents:

1. Colab - run it in your browser without installing anything locally
2. Setup - download pretrained models and resources
3. Pretrained Models - quickstart with pretrained models
4. Interactive Notebooks - jupyter notebooks for interactive navigation
5. Videos - export flying videos

• Layout Model
• Triplane Model

• reference):

git clone --depth 1 --filter=blob:none --sparse \
         https://github.com/google-research/google-research.git
cd google-research
git sparse-checkout set persistent-nature

• Install dependencies:
  • gcc-7 or above is required for installation. Update gcc following these steps.
  • We provide a Conda environment.yml file listing the dependencies. You can create a Conda environment with the dependencies using:

conda env create -f environment.yml

• Apply patch files: we provide a script for downloading associated resources applying the patch files

bash patch.sh

• Download the pretrained models:

bash download.sh

Stylegan3 (license), GSN (license), EG3D (license). Remaining changes are covered under Apache License v2

Contact

For any questions related to our paper, please email lrchai@mit.edu.