PERF: Panoramic Neural Radiance Field from a Single Panorama

Technical Report 2023

Guangcong Wang*¹ Peng Wang*² Zhaoxi Chen¹ Wenping Wang² Chen Change Loy¹ Ziwei Liu¹

S-Lab, Nanyang Technological University¹, The University of Hong Kong²

* denotes equal contribution

Project | YouTube | arXiv

Usage

Setup

Step 1: Clone this repository

git clone https://github.com/perf-project/PeRF.git
cd PeRF
pip install -r requirements.txt

Step 2: Install tiny-cuda-nn

pip install ninja
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

Step 3: Download checkpoints as shown here.

Train

Here is a command to train a PeRF of an example data:

python core_exp_runner.py --config-name nerf dataset.image_path=$(pwd)/example_data/kitchen/image.png device.base_exp_dir=$(pwd)/exp

Render a video

After training is done, you can render a traverse video by running the following command:

python core_exp_runner.py --config-name nerf dataset.image_path=$(pwd)/example_data/kitchen/image.png device.base_exp_dir=$(pwd)/exp mode=render_dense is_continue=true

Citation

Cite as below if you find it helpful to your research.

@article{perf2023,
    title={PERF: Panoramic Neural Radiance Field from a Single Panorama},
    author={Guangcong Wang and Peng Wang and Zhaoxi Chen and Wenping Wang and Chen Change Loy and Ziwei Liu},
    journal={Technical Report},
    year={2023}}