ggsonic / IRON

Inverse rendering by optimizing neural SDF and materials from photometric images

IRON: Inverse Rendering by Optimizing Neural SDFs and Materials from Photometric Images

Note: this repo is still under construction.

Project page: https://kai-46.github.io/IRON-website/

Usage

Create environment

git clone https://github.com/Kai-46/iron.git && cd iron && . ./create_env.sh

Download data

. ./download_data.sh

Training and testing

. ./train_scene.sh drv/dragon

Once training is done, you will see the recovered mesh and materials under the folder ./exp_iron_stage2/drv/dragon/mesh_and_materials_50000/. At the same time, the rendered test images are under the folder ./exp_iron_stage2/drv/dragon/render_test_50000/

Citations

@inproceedings{iron-2022,
  title={IRON: Inverse Rendering by Optimizing Neural SDFs and Materials from Photometric Images},
  author={Zhang, Kai and Luan, Fujun and Li, Zhengqi and Snavely, Noah},
  booktitle={IEEE Conf. Comput. Vis. Pattern Recog.},
  year={2022}
}

Example results

dragon.mp4

Acknowledgements

We would like to thank the authors of IDR and NeuS for open-sourcing their projects.