CityStreet: Multi-View Crowd Counting Dataset

The multi-view crowd counting datasets, used in our “wide-area crowd counting” paper, include our proposed dataset CityStreet, as well as two existing datasets PETS2009 and DukeMTMC repurposed for multi-view crowd counting.

The datasets files released consist of an instruction document and the following folders for each dataset:

1) Image frames:

  • CityStreet frames are provided; For PETS2009 and DukeMTMC, the frames can be downloaded or obtained by following the instructions in the document.

2) Labeling_tool:

  • A html-based multi-view labeling tool is provided for CityStreet and PETS2009 dataset. 

3) Labels:

  • The json files containing camera-view and ground-plane people labels.

4) Projection_code:

  • The python code for image2world and world2image projection.

5) GT_density_maps:

  • The ground-truth density maps, including both camera-view and scene-level density maps based on the corresponding labels.

6) ROI_maps and Metadata:

  • the camera-view and ground-plane ROIs of each view,
  • the normalization maps used in the late fusion,
  • the weighted maps used in the comparison method 1 “dmap_weighted”,
  • the distance maps used for scale selection in the MVMS model,
  • and the scene map used for visualization.

The datasets files links:

The dataset is about 16 GB. You can download it here:

7-zip should be able to extract the files of CityStreet; Linux users can use the following commands to combine and unzip multiple files:

    zip -s 0 CityStreet.zip --out single.zip
    unzip single.zip

If you use the dataset or the labels, please remember to cite our paper: 

  • Wide-Area Crowd Counting via Ground-Plane Density Maps and Multi-View Fusion CNNs.
    Qi Zhang and Antoni B. Chan,
    In: IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), Long Beach, June 2019. [dataset]
  • @inproceedings{zhang2019wide,
    title={Wide-Area Crowd Counting via Ground-Plane Density Maps and Multi-View Fusion CNNs},
    author={Zhang, Qi and Chan, Antoni B},
    booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
    pages={8297–8306},
    year={2019}
    }

If you have questions, contact qzhang364@my.cityu.edu.hk.

Acknowledgements: 

This work was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (CityU 11212518).