Skip to content

Latest commit

 

History

History
46 lines (31 loc) · 1.2 KB

README.md

File metadata and controls

46 lines (31 loc) · 1.2 KB

Frequency-Aligned Supervision for Few-shot Neural Rendering

Installation

We use cuda-11.1 in our experiments.

conda create -n p2nerf python=3.6.15
conda activate p2nerf
pip install --upgrade pip
pip install --upgrade jaxlib==0.1.68+cuda111 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
pip install -r requirements.txt

Data

You can download the data we processed from here and unzip the downloaded zip into the data/ folder.

Keypoints Prior [Optional]

If you want to generate keypoint prior data, you first need to install torch, plyfile and kornia, and then run the get_kpts_prior.py script.

# install
pip install torch==1.10.0+cu111 -f https://download.pytorch.org/whl/torch_stable.html
pip install kornia plyfile
# generate keypoints prior
python get_kpts_prior.py --data_dir data/P2NeRF/DDP --out_dir data/P2NeRF/prior2/DDP

Running

Scannet dataset

chmod +x ./scripts/ddp4.sh
./scripts/ddp4.sh

Acknowledgments

This code heavily references the P2NeRF codebases, and the authors are thanked for their open source behaviour.