Description

• This is the open source implementation of the CVPR2022 paper "Iterative Deep Homography Estimation"
  Si-Yuan Cao, Jianxin Hu, ZeHua Sheng, Hui-Liang Shen
  Zhejiang University, Hangzhou, China
• Paper link
• An improved version of IHN will be released soon.

Table of Contents

• Requirements
• Usage
• License
• Citation
• Contact
• Acknowledgement

Requirements

• Create a new anaconda environment and install all required packages before running the code.

  conda create --name ihn
  conda activate ihn
  pip install requirements.txt

Usage

The configurations are made in file .vscode/launch.json.
Please notice the location of the datasets in file datasets_4cor_img.py (for mscoco and spid) and datasets.py(for googlemap and googleearth).
Please note that the json file is only tested in vscode terminal.

License

This project is released under the Apache 2.0 license.

Citation

@InProceedings{Cao_2022_CVPR,
    author    = {Cao, Si-Yuan and Hu, Jianxin and Sheng, Zehua and Shen, Hui-Liang},
    title     = {Iterative Deep Homography Estimation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {1879-1888}
}

Contact

• Email: karlcao@hotmail.com

Acknowledgement

• This work is mainly based on RAFT and SCV, we thank the authors for the contribution.