Stay up to date with notifications from The Independent

Notifications can be managed in browser preferences.

Microsoft Project Oxford: New site launched that can recognise how people are feeling in photos

Fresh from guessing everyone’s age, and often getting it wrong, the site is having a go at emotions

Andrew Griffin
Thursday 12 November 2015 10:36 GMT
Comments
Does this woman look happy, or sad? Or maybe even angry. Or confused.
Does this woman look happy, or sad? Or maybe even angry. Or confused. (Shutterstock)

Microsoft has unveiled a new robot that can tell how people are feeling.

The algorithm can analyse a photo and attempt to pick out all of the emotions that are displayed in the face. It works in the same way as a tool that can guess the age of people in a picture, which was released earlier this year.

Project Oxford can be accessed through a special Microsoft site. That site also shows some of the process that the algorithm uses — first picking out where the face is, and then scoring the picture for various different emotions, such as anger, contempt, disgust and fear.

Microsoft’s description says that the tool “uses world-class machine learning techniques” to “identify emotions communicated by the facial expressions in an image”. It refers to the feature as Emotion API.

In use, the tool isn’t always accurate. It works far better on simple, obvious facial expressions (like a picture of someone with a huge grin, or their mouth open in surprise) than it does on subtle or mixed ones.

It isn’t clear how Microsoft intends to use the emotional recognition tool in its serious and mainstream products, but it’s easy to imagine how it would be useful on social networks or in search engines. The age-guessing tool eventually found its way into Bing’s image search, where a popup appears on pictures of people allowing users to click it and see how old the robot thinks the person is.

Join our commenting forum

Join thought-provoking conversations, follow other Independent readers and see their replies

Comments

Thank you for registering

Please refresh the page or navigate to another page on the site to be automatically logged inPlease refresh your browser to be logged in