top of page

We collected and annotated 60,000 frames

with rich information, leading to about 900,000 annotated objects.

You can see the the sample annotation on the right

The dataset is released under Creative Commons Attribution 4.0 License and our user agreement.


You can try out our sample data by click the button below. it contains data from one camera with three different time window.

Full Dataset download: By clicking on the Download button, you agree to our License and our user agreement. If you have any questions regarding the dataset, please contact us at

All submitted papers or any publicly available text using the CityCam database must cite the following paper:

Shanghang Zhang, Guanhang Wu, João P. Costeira, and José MF MouraUnderstanding Traffic Density from Large-Scale Web Camera Data, In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 4264-4273.

Related papers:

H Zhao, S Zhang, G Wu, JP Costeira, JMF Moura, GJ Gordon, Adversarial Multiple Source Domain Adaptation, Neural Information Processing Systems (NIPS) 2018

S Zhang, G Wu, JP Costeira, JMF Moura, FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras, International Conference on Computer Vision (ICCV) 2017

bottom of page