This is the official repository of the paper:

IDDR-NGP:Incorporating Detectors for Distractors Removal with Instant Neural Radiance Field

Install

Install with pip

pip install -r requirements.txt

Install with conda

conda env create -f environment.yml
conda activate iddr-ngp

Data

We use the same data format as instant-ngp. We provide three real scenes mentioned in the paper, please download and put them under ./data.

Usage

# use the 'outdoor' dataset
python main_nerf.py data/outdoor --workspace outdoor_test -O --demask

Result

The processed image will be saved in the ./your_workplace/result/.

Citation

If you find this work useful, a citation will be appreciated via:

@inproceedings{huang2023iddr,
  title={IDDR-NGP: Incorporating Detectors for Distractors Removal with Instant Neural Radiance Field},
  author={Huang, Xianliang and Gou, Jiajie and Chen, Shuhang and Zhong, Zhizhou and Guan, Jihong and Zhou, Shuigeng},
  booktitle={Proceedings of the 31st ACM International Conference on Multimedia},
  pages={1343--1351},
  year={2023}
}

Acknowledgement

The implementation is based on the torch-ngp.

Contact

If you have any question, please feel free to contact huangxl21@m.fudan.edu.cn.