This is the official reference implementation of the paper "NeRFCodec: Neural Feature Compression Meets Neural Radiance Fields for Memory-Efficient Scene Representation."
Our work builds upon excellent open-source contributions in the NeRF and Neural Compression fields:
Please adhere to their licenses. We thank the authors of these outstanding works and their repositories.
First, install environment (which is basically same as TensoRF):
conda create -n NeRFCodec python=3.8
conda activate NeRFCodec
pip install torch==2.2.1 torchvision==0.17.1 --index-url https://download.pytorch.org/whl/cu118
pip install tqdm scikit-image opencv-python configargparse lpips imageio-ffmpeg kornia lpips tensorboard
pip install matplotlib plyfile pytorch_msssim
Single scene training
- Pre-train the TensoRF with 30,000 iterations:
bash scripts/tensorf_train/run_chair.sh- Warm up the modified neural codec and proceed with the joint training of the neural feature codec and the neural radiance field for 100,000 iterations:
bash scripts/joint_training/run_chair_feat_codec.sh # low rate point
bash scripts/joint_training/run_chair_feat_codec_384.sh # high rate pointMultiple scene parallel training We reuse the training script "auto_run_paramsets.py" from TensoRF project for the execution of parallel training.
To be released...
If you have any issue regrading this repository, e.g. environment setup, training scripts, more technical details, please e-mail Sicheng Li at [email protected].
If you find our code or paper useful, please cite
@InProceedings{Li_2024_NeRFCodec,
author = {Li, Sicheng and Li, Hao and Liao, Yiyi and Yu, Lu},
title = {NeRFCodec: Neural Feature Compression Meets Neural Radiance Fields for Memory-efficient Scene Representation},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2024}
}