Microsoft Research peels back the curtain on its viral hit, How-Old.net, which uses a machine-learning technology called Project Oxford. HoloLens, Microsoft's buzz-worthy augmented-reality technology, wasn't the only thing that resonated with the IT community following last week's Build developer conference.
How-Old.net, a Website where users upload photos and it guesses their age and gender, was a hit on social media and attracted widespread tech coverage over the weekend. Once photos are uploaded, the site draws a box around the subjects' faces, along with their ages and a male or female icon.
Just three hours after the team sent an internal email, users flooded the Internet with screenshots of their supposed age, which ranged from spot-on to humorously inaccurate.
"Within hours, over 210,000 images had been submitted and we had 35,000 users from all over the world (about 29K of them from Turkey, as it turned out—apparently there were a bunch of tweets from Turkey mentioning this page)," Corom Thompson and Santosh Balasubramanian, engineers in Information Management and Machine Learning at Microsoft, wrote in a company blog post.
Predictably, many users uploaded images of celebrities and other recognizable people. "But over half the pictures analyzed were of people uploading their own images," said Thompson and Balasubramanian. "This insight prompted us to improve the user experience and we did some additional testing around image uploads from mobile devices."
The site is based, in part, on the face-recognition component of Project Oxford, a collection of Azure machine-learning application programming interfaces (APIs) and services currently in beta. "This technology automatically recognizes faces in photos, groups faces that look alike and verifies whether two faces are the same," Allison Linn, a Microsoft Research writer, stated in a separate blog post.
Apart from guessing ages, Linn noted that the technology has other, potentially more business-friendly applications. "It can be used for things like easily recognizing which users are in certain photos and allowing a user to log in using face authentication."
Thompson and Balasubramanian admitted that How-Old.net may miss the mark. "Now, while the API is reasonably good at locating the faces and identifying gender, it isn't particularly accurate with age, but it's often good for a laugh and users have fun with it."
Microsoft is increasingly relying on its machine-learning research to enhance its software and services portfolio. "We want to have rich application services, in particular, data services such as machine learning, and democratize the access to those capabilities so that every developer on every platform can build intelligent apps," said CEO Satya Nadella during his opening remarks at Build.
In February, Microsoft announced the general availability of its cloud-based predictive analytics offering, Microsoft Azure Machine Learning. T. K. Rengarajan, corporate vice president of the Data Platform unit, and Joseph Sirosh, corporate vice president of Machine Learning said in a statement at the time that "developers and data scientists can build and deploy apps to improve customer experiences, predict and prevent system failures, enhance operational efficiencies, uncover new technical insights, or a universe of other benefits" with the big data processing platform in mere hours.