Face recognition is being increasingly used for all sorts of things, but now you can add age estimation into the mix. Just by analysing a photo of your face, the software can estimate your age and place an upper and lower bound on the possible range.
Face.com provides a (currently free) face recognition API that you can sign up to and use. The API will already detect a face, tell you the sex of the person, if they are wearing glasses and estimate one of a number of possible mood categories - happy, sad, etc. Now it will also provide an estimate of how old the person is.
The API will return three integers per face - minimum age, maximum age and estimated age - along with a confidence estimate for each. You can also set an age range for faces that you want detected within photos. So for example you could ask for all of the people aged between 18 and 25 to be detected.
No information on how the algorithm works has been provided, but it isn't difficult to guess at the sorts of features that are age related. Apparently the estimator was trained on thousands of sample pictures. The target age used in the training was estimated by humans rather than actual chronological age. The means that the system is actually attempting to estimate apparent age. It also suggests that accuracy might vary according to how easy it is to guess a persons age from a photo e.g. some people show their age, others look timeless.
If you want to try it out there is a test page that you can play with and even upload your own photos to test.
What could you use this for?
The Face.com blog has some suggestions:
The uses of this capability are nearly limitless, from parental control applications (restricting or enabling age-limited functionality based on the visitor’s age) to potential real-world targeted advertising based on detected age range of a consumer.
There are, of course limitless possibilities if you want to get a little creative - for example what about comparing apparent age to chronological age as a way to tell the user how well they are standing up to the ravages of time? Of course this might be a step too far.
Face.com also claims to have improved the accuracy of its recognition engine by 30% and improved its service by moving to a new data center.
The package, which is now on GitHub, is written in R and automatically detects anomalies such as spikes in data, which happen on Twitter when a major news item breaks, or there's a major sportin [ ... ]