This is the official repo for the implementation of Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction, Qiancheng Fu*, Qingshan Xu*, Yew-Soon Ong, Wenbing Tao (* Equal Contribution), NeurIPS 2022.
We will release our pretrained models soon!
This code is built with pytorch 1.11.0 and pytorch3d 0.6.2. In addition, other packages listed in requirements.txt are required.
You can create an anaconda environment called geoneus with the required dependencies by running:
conda create -n geoneus python=3.7
conda activate geoneus
conda install pytorch==1.11.0 torchvision==0.12.0 cudatoolkit=11.3 -c pytorch
conda install fvcore iopath
conda install -c bottler nvidiacub
conda install pytorch3d -c pytorch3d
pip install -r requirements.txt
Download the DTU data from Google Drive, which provides the files on sparse points and image pairs.
- Training
python exp_runner.py --mode train --conf ./confs/womask.conf --case <case_name>
- Extract surface from trained model
python exp_runner.py --mode validate_mesh --conf ./confs/womask.conf --case <case_name> --is_continue
- Evaluation
python eval.py --conf ./confs/womask.conf --case <case_name>
If you find our work useful in your research, please consider citing:
@article{Fu2022GeoNeus,
title={Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction},
author={Fu, Qiancheng and Xu, Qingshan and Ong, Yew-Soon and Tao, Wenbing},
journal={Advances in Neural Information Processing Systems (NeurIPS)},
year={2022}
}
Our code is partially based on NeuS project and some code snippets are borrowed from NeuralWarp. Thanks for these great projects.