PanoIndoor

Indoor large-scale panoramic visual localization dataset
Hailin Yu, Jiarun Liu, Zhichao Ye, Xinyu Chen, Ruohao Zhan, Yichun Shentu, Zhongyun Lu, Guofeng Zhang
JIG 2025

Dataset Link

通过网盘分享的文件：PanoIndoor
链接: https://pan.baidu.com/s/1JUtCXu6yTr0rJaFPCFrGkA?pwd=kp54 提取码: kp54 
--来自百度网盘超级会员v7的分享

Dataset Structure

Our dataset contains four scenes:

• office
• canteen
• corridor
• parkinglot

Each scene contains files in following structure:

.
├── images                      # all panorama images
│   └── insta_pano_default
├── query_with_intrinsics.txt   # query images with pose
└── sfm_workspace               # reconstructed map with all images
    ├── cameras.bin
    ├── images.bin
    └── points3D.bin

Tools

We also provide three useful tools for data process:

• tools\face_mask is for data desensitization. We have mosaic processed the visible faces in the dataset. If there are more advanced privacy protection requirements, manual processing can be performed.
• tools\map_split is for spliting the maps in the sfm_workspace based on the selected reference image folder name. During localizaiton, we only need the reconstruction results of the references, but not the queries.
• tools\panorama is for changing panorama images to perspective images.
  Please check README.md in each folder for detailed usage.

Citation

If you find this code useful for your research, please use the following BibTeX entry.