JHU-CROWD++: A large-scale unconstrained crowd counting dataset

A comprehensive dataset with 4,372 images and 1.51 million annotations. In comparison to existing datasets, the proposed dataset is collected under a variety of diverse scenarios and environmental conditions. In addition, the dataset provides comparatively richer set of annotations like dots, approximate bounding boxes, blur levels, etc.

Dataset Highlights

Includes several images with weather-based degradations and illumination variations, making it a very challenging dataset. Additionally, the dataset consists of rich annotations at both image-level and head-level.

  • 4,372 images

    Contains 4,372 images (with an avg resolution of 1430x910) collected under a diverse set of conditions and various geographical locations.

  • Diverse conditions

    Specific care is taken to improve diversity of the dataset by including images under adverse weather and various illumination conditions.

  • 1.51 million annotations

    Contains a total of 1.51 million dot annotations with an average of 346 dots per image and a maximum of 25K dots.

  • Rich set of annotations

    Provides head-level labels (dots, approx. bounding box, blur-level, etc.) and image-level labels (scene type and weather condition).

Image Highlights

Diverse Conditions: varying densities, illumination variations, adverse weather conditions such as fog, rain and snow.

Foggy Photo Dataset

Low density

Foggy Photo Dataset

Medium Density

Foggy Photo Dataset

High Density

Rainy Photo Dataset

Fog

Rainy Photo Dataset

Rain

Rainy Photo Dataset

Snow

Snowy Photo Dataset

Low illumination

Snowy Photo Dataset

Night time

Snowy Photo Dataset

Distractor

Annotation Highlights

Rich set of annotations: dots, approximate bounding boxes, blur-levels, etc.

Dots

Bounding Box

Blur-level

Dots

Bounding Box

Blur-level

Dots

Bounding Box

Blur-level

Statistics

Distribution of image labels

Distribution of different density images

Distribution of weather-degradations

Citation

If you find this dataset useful, please consider citing the following work:

@inproceedings{sindagi2019pushing,
title={Pushing the frontiers of unconstrained crowd counting: New dataset and benchmark method},
author={Sindagi, Vishwanath A and Yasarla, Rajeev and Patel, Vishal M},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
pages={1221--1231},
year={2019}
}
@article{sindagi2020jhu-crowd++,
title={JHU-CROWD++: Large-Scale Crowd Counting Dataset and A Benchmark Method},
author={Sindagi, Vishwanath A and Yasarla, Rajeev and Patel, Vishal M},
journal={Technical Report},
year={2020}
}

Acknowledgements

We would like to express our deep gratitude to everyone who contributed to the creation of this dataset including the members of the JHU-VIU lab and the numerous Amazon Mturk workers. We would like to specially thank Kumar Siddhanth, Poojan Oza, A. N. Sindagi, Jayadev S, Supriya S, Shruthi S and S. Sreevali for providing assistance in annotation and verification efforts.

Lastly, we would like to thank Kannan Kandappan for the landing page design.

Contact

Send an email to vishwanathsindagi@jhu.edu for any queries.

Ready to dive in?
Download this dataset today.