Official implementation of AAAI 2025 paper "Object-level Geometric Structure Preserving for Natural Image Stitching".

1. Compile Opencv 4.4.0, VLFEAT and Eigen locally.

2. Create a new Visual Studio .sln, and add all .cpp and .h files into this .sln.

3. Set HED file paths in EdgeDetection.cpp.

StitchBench is by far the most comprehensive image stitching dataset. A sample image pair is provided in ./input-data/AANAP-01_skyline. StitchBench is available at: HuggingFace. You will be automatically granted access to it.

Aerial234 is a open-source dataset of 234 aerial images for image stitching. We used a drone to continuously scan an area of Southeast University and collected this dataset. It’s quite a challenging dataset, and we’re curious if there’s a method to stitch these 234 aerial images into a single panorama.

Aerial234 is available at: HuggingFace.

Our work on aerial image stiching (just a preliminary attempt): UAV image stitching by estimating orthograph with RGB cameras.

Run .sln and you will find 0-original.png in the ./ folder. Upload the image to Google Colab and run sam.ipynb to get SAM features and put it in ./ folder.

For any questions, please feel free to open an issue.

@inproceedings{cai2025object,
  title={Object-level geometric structure preserving for natural image stitching},
  author={Cai, Wenxiao and Yang, Wankou},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={39},
  number={2},
  pages={1926--1934},
  year={2025}
}

@article{cai2023uav,
  title={UAV image stitching by estimating orthograph with RGB cameras},
  author={Cai, Wenxiao and Du, Songlin and Yang, Wankou},
  journal={Journal of Visual Communication and Image Representation},
  volume={94},
  pages={103835},
  year={2023},
  publisher={Elsevier}
}

We appreciate AAAI for providing Student Scholarship for this paper!