Xiaolin Wu and Xi Zhang, two researchers of the University of Shanghai Jiao Tong (Chinese), just of publish a paper in which is discussed them results retrieved through a project where is has created a network neuronal capable of identify criminal by their features facial attending to a premise somewhat peculiar as is that all and every one of them share a series of patterns physical similar.
This idea that criminals share similar physical patterns was born in the 19TH century, shortly after the invention of photography, although curiously it was not until the middle of the year 2011 when could be demonstrated with certainty that it is possible to differentiate those who violate law which meet it just to see his photographs.
Create a neural network capable of identifying, with 89.5% success probability, criminals by their facial features.
With this idea in mind, is much more simple of understand the base so these researchers Chinese have got develop, using to its time a wide variety of algorithms of vision, a network neuronal trained starting from them images of them faces of 1.856 men Asian of between 18 and 55 years, shaved and without brands facial where 730 were criminal that currently is found fulfilling condemns or in whereabouts unknown. Among the 730 criminal is included the images of 235 accused of crimes violent.
Once trained neural network with 90% of the total of images was used 10% of the remaining images to test the obtained result. The result was that this new software was able to correctly identify the criminals with a precision of 89.5%. According to the authors of this work, the neural network is based on three main facial argos to identify a criminal, such as the curvature of the upper lip, the distance between the bottom corners of the eyes and the angle between two lines drawn from the tip of the nose to the corners of the mouth.
More information: MIT