Google brings photo and video databases for machine learning




Google has released two databases, one with video data, and another one with image data. The data should help researchers in training algorithms for machine learning systems.

The database with pictures is the Open Images Dataset called , and consists of nine million images that have been tagged. These tags should help self-learning algorithms to recognize images. Because there are six thousand are several categories, systems need to learn to recognize a great diversity of images.

Google has initially uses a proprietary algorithm to tag the photos, but the validation is done by people. To be able to Open Images has the Internet giant collaborated with Cornell University and Carnegie Mellon University.

Google Open Images

Earlier this week released Google already has a different database from the YouTube 8M. As the name implies, this dataset includes eight million videos from YouTube. Just like in the Open Database Images are videos tagged, making algorithms can train themselves to recognize video images.

In total, the dataset consists of half a million hours of video and 1.9 billion frame features. In addition, there are 4,800 different kinds of videos so that researchers can train their algorithms with a wide variety of video content. There are only YouTube videos used with more than one thousand views; according to Google must guarantee adequate quality.

Google argues that the release of the data sets in particular can help investigators. They often do not have access to large archives of images to train their machine learning algorithms. The Internet giant hopes the data sets to ensure that more research is done.


In: A Technology & Gadgets Asked By: [21995 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »