Google acquires DNNresearch (deep learning/image search)
Companies in Relationship
Initial Announcement Date:
Consideration: not disclosed
Google acquires DNNresearch, which developed a system for image search based on deep learning and convolutional neural networks. The technology was developed by Professor Geoffrey Hintonís group at the University of Toronto. Google had granted $600K to the U. of Toronto research team.
Googleís previous photo search relied primarily on image filenames and text from the web to classify images. This presented challenges with searching personal photos since they generally do not have filenames or information that provide context for their content. A group at Google was already working to develop a photo search system using a similar approach to Hintonís group. When DNNresearch won the ImageNet computer vision competition by a wide margin, Google realized that DNNesearchís technology would aid its efforts to make photo search more precise and easier. As mentioned above, Google had previously backed Hinton and his teamís research with a $600K grant, but the acquisition allowed company to adopt Hintonís technology and guide his teamís research more directly.
From a competitive standpoint Microsoft had already rolled out deep neural network technology in some of its audio- and video-indexing systems. Googleís acquisition helped the firm keep pace in this quickly developing area.
DNN's technology is now used for photo search in Google+, with the enhanced service being developed in about six months.
Copyright © 2008-2015 George T. Geis. All Rights Reserved.