Let’s picture the scene, you upload a snap of a special occasion to your Facebook account to share with your friends, family and followers, as there are many people in the scene you decide to tag everyone individually, you think this will be easier for everyone, but there’s a problem, Facebook doesn’t recognize you as the moment the flash went off, you were pictured with your head to one side attempting to avoid a wasp which came hurtling towards you at warp speed.
Help may soon be at hand with an experimental algorithm which has been devised by Facebook’s artificial intelligence lab. This new technique is software which can recognise people in photographs even when it is unable to see their faces; instead it relies on points of reference which are unique to each individual. This can include hair style, specific clothing and even the pose which you often strike.
In order to test the final algorithm, researchers downloaded 40,000 public photos from Flickr which contained a mixture of clearly and obscure images of individuals. The algorithm was able to recognize the identities of individuals with an 83% success rate. As this is Facebook I am sure privacy of these photos were high on their agenda, cough cough.
This experiment does have the potential for real world applications, but there needs to be caution exercised if a machine can identify you without you wanting it to. If someone takes a photo before uploading the image to Facebook and you happen to be in the background, this algorithm will identify you for all to see.
Thank You New Scientist for providing us with this information
Image Courtesy of Geek Snack