Doubles and three-dimensional digital clones, a reality soon
From the University of Washington a method for creating animated three-dimensional models of any individual starting from a set of photographs not of high quality.
It may soon be possible to create digital doppelganger of anyone simply collecting any type of image spread on the network and create a sort of ” digital puppetry ” that may take any action and say anything. Researchers at the University of Washington have developed a METHOD; that allows to create digital elevation models starting from any set of photographs of the same subject, such as those that may be found on the Internet.
Supasorn Suwajanakorn, coordinator of the study, says: ” It may be relatively easy in the future to create digital models of anyone when there are tons of photos. The photos on Facebook, for some users, are already enough to build their model controllable “. The study, funded by Samsung, Google and Intel, along with the University of Washington, will be detailed at the Conference on Computer Vision in Chile on December 16th.
Existing technologies for creating three-dimensional models detailed in human beings lean complex methods to capture their movements from every point of view. The researchers instead found a way to create templates that you can animate based solely on a random collection of images pre-existing and have been able to reconstruct the movements and facial appearance of celebrities like Daniel Craig and Neil Patrick Harris or politicians like George W. Bush and Barack Obama using 200 images of each retrieved Web, taken over the years in various poses and settings.
The method produces three-dimensional model of an individual making an ” average ” of the images analyzed. The team then developed a series of methods that allow you to modify the expression of the model in a subtle way, to reproduce the characteristics of a grimacing smile or various reactions and emotional manifestations, also tracing the movements of another individual.
It is a method that has proven its effectiveness with any expression, in any installation and in various lighting conditions, or the variety of photographs that you can expect to find in the personal profile of the majority of users Facebook. Suwajanakorn explains: ” Unlike other techniques of 3D reconstruction is not necessary that the subject is before the camera, which offers the potential to scale on a large number of individuals. ”
The technology currently only works on pictures of faces: ” We have no voice, the rest of the head or specific ways that a person does certain things. It is more difficult to collect, especially for ordinary people, and raise some problems privacy “.
This technology could be used, for example, in combination with virtual reality devices or increased, to make more engaging video calls and video conferences, or in the field of gaming to interact with celebrities. But researchers are also aware of the possible dangers involved: ” It could be quite scary when rendering will be indistinguishable from the real subject. People will begin to consider the video carefully, as if it has become easy to sweeten a photo with Photoshop. “
