The WWW dataset provides 10,000 videos with over 8 millionframes from 8,257 diverse scenes, therefore offering a superiorlycomprehensive dataset for the area of crowd understanding. The abundantsources of these videos also enrich the diversity and completeness.Related publications:Jing Shao, Kai Kang, Chen Change Loy, and Xiaogang Wang. "Deeply learned attributes for crowded scene understanding". in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (PDF)