Skip to content
/ GPF Public

[ICLR2024] Official implementation of Generalizable Point Field

License

Notifications You must be signed in to change notification settings

Mercerai/GPF

Folders and files

NameName
Last commit message
Last commit date

Latest commit

author
mercer
Jun 14, 2024
4b47b3f · Jun 14, 2024

History

20 Commits
Jun 12, 2024
Jun 13, 2024
Jun 13, 2024
Jun 12, 2024
Jun 13, 2024
Jun 13, 2024
Jun 12, 2024
Jun 12, 2024
Jun 12, 2024
Jun 12, 2024
Jun 12, 2024
Jun 14, 2024
Jun 12, 2024
Jun 12, 2024
Jun 12, 2024
Jun 13, 2024
Jun 12, 2024
Jun 12, 2024

Repository files navigation

GPF: Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation

Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation
Jiaxu Wang, Ziyi Zhang, Renjing Xu*
ICLR 2024

framework_img

If you found this project useful, please cite us in your paper, this is the greatest support for us.

Installation

git clone https://github.com/Mercerai/GPF.git
cd GPF
pip install -r requirements.txt
Dependencies
  • torch==1.7.1
  • numpy==1.19.2
  • CUDA 11.4 or later version

Dataset

We reorganize the original datasets in our own format. Here we provide a demonstration of the test set of DTU, which can be downloaded here. After placing the demo data into the data directory, one can directly run the test code as follows. In the data_preprocess dir, we provide the code to reorganize the original datasets into our format.

Pretrained models

We provide the pretrained model which can be applied to the DTU and BlendedMVS datasets.

Evaluation for rendering a novel camera trajectory

python test.py --config ./configs/dtu_config --render_path --num_views 15 --interp_0 14 --interp_1 16 --scan_num 114

Evaluation for rendering a certain camera viewpoint in the test set

python test.py --scan_num 114 --view_num 36

--render_path switch on if camera path is rendered
--num_views how many frames will be rendered on the path
--interp_0 and 1 frames between which need to be interpolated
--scan_num the scan number
--view_num specific view number

The results will be saved in ./log/dtu_eval/

Acknowledgements

In this repository, we have used codes or datasets from the following repositories. We thank all the authors for sharing great codes or datasets.

Citation

@inproceedings{jiaxu2023learning,
  title={Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation},
  author={Jiaxu, WANG and Zhang, Ziyi and Xu, Renjing},
  booktitle={The Twelfth International Conference on Learning Representations},
  year={2024}
}

About

[ICLR2024] Official implementation of Generalizable Point Field

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published