At the IEEE Conference on Computer Vision and Pattern Recognition in June 2009, Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei presented “ImageNet: A Large-Scale Hierarchical Image Database.” The publisher record confirms the title, the six authors, the CVPR 2009 venue, and the June 2009 date. The official dataset site at image-net.org describes the resource as an image database organized by the WordNet hierarchy, with millions of labeled images, built at Stanford and Princeton.
The contribution was data, not a new algorithm. By assembling labeled images at a scale far beyond anything before it, the team gave researchers a common, hard benchmark to measure progress against. The annual ImageNet competition turned that benchmark into a yearly contest.
ImageNet proved decisive. In 2012 a deep convolutional network called AlexNet won the ImageNet competition by a wide margin, an event that convinced the field that deep learning plus large data plus GPUs was the winning formula. ImageNet is the clearest case in this era of a dataset shaping the direction of an entire field.