CrowdHuman

A Benchmark for Detecting Human in a Crowd

Shuai Shao*, Zijian Zhao*, Boxun Li, Tete Xiao, Gang Yu, Xiangyu Zhang, Jian Sun

CrowdHuman is a benchmark dataset to better evaluate detectors in crowd scenarios. The CrowdHuman dataset is large, rich-annotated and contains high diversity. CrowdHuman contains 15000, 4370 and 5000 images for training, validation, and testing, respectively. There are a total of 470K human instances from train and validation subsets and 23 persons per image, with various kinds of occlusions in the dataset. Each human instance is annotated with a head bounding-box, human visible-region bounding-box and human full-body bounding-box. We hope our dataset will serve as a solid baseline and help promote future research in human detection tasks.

News

[news] Our dataset is published.

fig1.png

Volume, density and diversity of different human detection datasets.
For fair comparison, we only show the statistics of training subset.

fig2.png

Comparison of the visible ratio between our CrowdHuman and CityPersons dataset.
Visible Ratio is defined as the ratio of visible bounding box to the full bounding box.

examples
examples
examples

examples
examples
examples

examples
examples
examples

Challenge

Coming Soon...

Paper

paper

Download

download

Please cite the following paper if you use our dataset.


  @article{shao2018crowdhuman,
    title={CrowdHuman: A Benchmark for Detecting Human in a Crowd},
    author={Shao, Shuai and Zhao, Zijian and Li, Boxun and Xiao, Tete and Yu, Gang and Zhang, Xiangyu and Sun, Jian},
    journal={arXiv preprint arXiv:1805.00123},
    year={2018}
  }
          
shaoshuai

Shuai Shao

Megvii Inc.

zhaozijian

Zijian Zhao

Beihang University

liboxun

Boxun Li

Megvii Inc.

xtt

Tete Xiao

Peking University


yugang

Gang Yu

Megvii Inc.

zhangxiangyu

Xiangyu Zhang

Megvii Inc.

sunjian

Jian Sun

Megvii Inc.