Datasets for image annotation and retrieval

This page hosts datasets used in our CSUR paper Socializing the semantic gap: A comparative Survey on Image Tag Assignment, Refinement and Retrieval.

Train10k Train100k Train1m MIRFlickr Flickr51 NUS-WIDE
Nr. of images 10,000 100,000 1,198,818 25,000 81,541 259,233
Nr. of test tags 14 51 81
vggnet16-fc7relu train10k-vggnet16-fc7relu.tar.gz train100k-vggnet16-fc7relu.tar.gz train1m-vggnet16-fc7relu.tar.gz mirflickr08-vggnet16-fc7relu.tar.gz flickr51-vggnet16-fc7relu.tar.gz flickr81-vggnet16-fc7relu.tar.gz
googlenet-pool5 in preparation in preparation in preparation in preparation in preparation in preparation
social tags train10k-tag.tar.gz train100k-tag.tar.gz train1m-tag.tar.gz mirflickr08-tag.tar.gz flickr51-tag.tar.gz flickr81-tag.tar.gz
tag frequency train10k-tagfreq.tar.gz train100k-tagfreq.tar.gz train1m-tagfreq.tar.gz  –  –  –
ground truth mirflickr08-anno.tar.gz flickr51-anno.tar.gz flickr81-anno.tar.gz
Flickr image urls train1m-urls.txt.gz nuswide

  • [Train1m] Xirong Li, Cees G. M. Snoek, Marcel Worring, and Arnold W. M. Smeulders, Harvesting social images for bi-concept search. IEEE Transactions on Multimedia, 14(4):1091-1104, 2012
  • [MIRFlickr] Mark J. Huiskes, Bart Thomee, and Michael S. Lew, New Trends and Ideas in Visual Concept Detection: The MIR Flickr Retrieval Evaluation Initiative, Proc. of ACM MIR, 2010
  • [NUS-WIDE] Tat-Seng Chua, Jinhui Tang, Richang Hong, Haojie Li, Zhiping Luo, and Yantao Zheng, NUS-WIDE: A real-world web image database from national university of singapore, Proc. of ACM CIVR, 2009
  • [Flickr51] Meng Wang, Kuiyuan Yang, Xian-Sheng Hua, and Hong-Jiang Zhang, Towards a relevant and diverse search of social images. IEEE Transactions on Multimedia, 12(8):829–842, 2010