Microsoft has taken the wraps off Azure Media Face Detector, a cloud service that can recognise people and determine their moods.
Redmond says the new clouds service “can be used for people counting, movement tracking, and even gauging audience participation and reaction via facial expressions.”
Microsoft's happy that the service can find up to 64 faces within a photo, and that even faces depicted at just 24x24 pixels can be identified as a human. 2048x2048 yields the best results and at that resolution Redmond is confident its analytical engines can deduce “multiple emotional attributes … including happiness, sadness, fear, anger, and more.”
Microsoft's actually a little bit late to this game. Call centres have used “sentiment analysis” for years to assess callers' moods on the phone. Microsoft's announcement launching the complete Azure Media suite mentions call centres as one target for its new cloudy voice analysis service. The company is also targeting video analysis because “With the growth of IP cameras, there is an explosion of surveillance videos. Manually reviewing surveillance video is time intensive and prone to human error.” Enter Azure Media Analytics “… to make the process of reviewing, managing and creating derivatives easier.”
Derivatives? That's where things might get scary if a user decided to mash up the output of Microsoft's services with their own store of known faces.
Microsoft's a clean-skin here: what Azure Media Face Detector users decide to get up to is their problem – and maybe ours – but not Microsoft's. The company has, however, made the output of the Face Detector available in lovely JSON, the better to facilitate easy consumption. ®