On a scale from one to
In a recent Amazon Web Services (AWS) post, which comes via CNBC, the behemoth corporation announced that its object recognition software, dubbed Rekognition, now has “improved accuracy for emotion detection,” which includes a spate of emotions like calm, confused, disgusted, surprised, angry, sad, and happy. And now added to that list: fear. In the post, Amazon tacked that one individually on the end of the list, noting that it’s a new identifiable emotion.
For those who haven’t heard of the Big Brother-esque Rekognition software, it’s basically a system of neural nets trained on big data (i.e.
While fear was specifically noted by Amazon as a new emotion that Rekognition can identify, it’s actually only one addition to a number of improvements the AI is making. Along with identifying emotions, Amazon says that Rekognition is now also better at spotting gender and age range. Combined with the aforementioned list of other objects that Rekognition can identify, it’s clear that Amazon wants this software to be able detect and label absolutely any type of image you throw at it.
There’s no explanation of how exactly Rekognition’s neural nets identify all these emotions, but obviously it’s spotting patterns in the way faces are contorted when reflecting different emotional states. Although, as CNBC notes, a lot of experts say that changes in a face don’t necessarily reflect a change in emotion — a pattern Rekognition could probably be able to verify or discard in the near future.
What do you think of Rekognition’s ability to identify emotions such as fear? Do you think facial contortions reflect emotions, or are you one of those people who always looks angry even when you’re happy? Let us know in the comments!