Codebase for Progressive Multi-Scale Light Field Networks (3DV 2022).
- Download our datasets and extract them to
datasets
directory. - Setup a PyTorch environment and install
requirements.txt
. - To train, run
python app.py -c configs/run_mipnet_jon.txt
.
Alternatively, download our trained LFNs toruns
.
To use the viewer on Ubuntu, run the following:
sudo apt install libmesa-dev libglfw3
# Required to install pycuda with OpenGL support
echo "CUDA_ENABLE_GL = True" > ~/.aksetup-defaults.py
pip install pycuda glumpy pyopengl
rm ~/.aksetup-defaults.py
python app.py -c configs/run_mipnet_jon.txt --script-mode=viewer
If you get CUBLAS_STATUS_EXECUTION_FAILED
while opening the viewer, try running with CUBLAS_WORKSPACE_CONFIG=:0:0
. (PyTorch Issue).
@inproceedings{li2022progressive,
author={Li, David and Varshney, Amitabh},
booktitle={2022 International Conference on 3D Vision (3DV)},
title={Progressive Multi-Scale Light Field Networks},
year={2022},
volume={},
number={},
pages={231-241},
doi={10.1109/3DV57658.2022.00035}}
}
utils/nerf_utils.py
is borrowed fromkrrish94/nerf-pytorch
.