Computer Graphics  

A newly developed technology that uses an inexpensive facial mapping method can recreate and replace faces of celebrities, but the technology may also have applications in creating digital avatars and mindclones.

Soon, nearly everyone with a computer will be able to create inexpensive, controllable computer models of famous people’s faces in 3D just using online photos of celebrities, or anyone.

"This capability opens up the ability to create puppets for any photo collection of a person, without requiring them to be scanned," write the researchers from the University of Washington.

The researchers also were able to recreate idiosyncrasies, like facial ticks, of the celebrities they modeled, including Kevin Spacey and Daniel Craig. This work could let us interact with digital avatars that look and act like people we know.

Related articles Researchers Rewire Paralyzed Monkey To Control Arm Directly From The Brain Infographic - Are We Living In A Computer Simulation? Is Brown Fat The Key To A Healthier You?

Typically, creating CGI models of a human face is expensive and requires laser scanning and motion capture technology. The researchers postulated that for celebrities, there are more than enough paparazzi photographs online to capture digitally what they look like from just about every possible angle.

"The idea was to create realistic virtual models of people just from photos rather than complex lab set-ups," says researcher Ira Kemelmacher-Shlizerman. Led by Supasorn Suwajanakorn, the team collected around 200 photos each of various famous people from Google Images, taken in different poses and at varying angles.

"Don't underestimate how much people want to be someone else."

The photos were analysed using face-tracking software, and a realistic 3D model for their face and head created. Further analysis added wrinkles and textures that appear and disappear as expressions change.

"The result is a full 3D model you can turn right around," says Suwajanakorn.

As well as being able to manipulate the digital puppets any way they wanted, the team found they could realistically switch an actor's face with that of another – allowing them to be replaced throughout an entire TV show or movie, for example.

A real potential commercial application could be to provide famous animated faces for visual versions of Siri-like assistants, says Suwajanakorn. Another might be to automatically create 3D faces for telepresence robots, says Nicole Carey at humanoid robot maker Engineered Arts in Penryn, UK. "Don't underestimate how much people want to be someone else," she says.

The face of someone who has died could be recreated and driven with one of the emerging breed of chatbots trained – using the deceased's tweets and emails – to converse like them.

The system may also be useful for creating mindclones. "Our model could bring back your memories of the people you care about," says Suwajanakorn.




SOURCE  New Scientist

By 33rd Square

Embed

 Neuroscience  

A new study has revealed that individual neurons in the human brain are triggered by the subject's conscious perception, rather than by the visual stimulus.

An international team of scientists has shown how individual neurons in the human brain react to ambiguous morphed faces. The study found neurons fire in line with conscious recognition of images rather than the actual images seen, and points to a clear role for neurons in the formation of memory.

The paper, ‘Single-Cell Responses to Face Adaptation in the Human Medial Temporal Lobe’, was published in the journal Neuron.

For this, the researchers including Professor Rodrigo Quian Quiroga, director of the Centre for Systems Neuroscience and Head of Bioengineering at the University of Leicester, used images of celebrities, such as Angelina Jolie and Halle Berry, morphed together to create an ambiguous face which test subjects were asked to identify.

Related articles Researchers Turn One Kind Of Neuron Into Another In A Living Brain  Dendrites Found To Process Information In the Brain  Human Memory Implants May Soon Be Tested

The study found that for the same ambiguous images, the neurons fired according to the subjective perception by the subjects rather than the visual stimulus. For example, a neuron originally firing to Whoopi Goldberg fired to a morph image between Goldberg and Bob Marley only when the subject identified the morphed image as Goldberg and remained silent when the subject said the very same image was Marley.

They concluded that neurons fire in line with conscious recognition of images rather than the actual images seen. Furthermore, in most cases the neuron’s responses to morphed pictures were the same as when shown the pictures without morphing.

The study was carried out by Quiroga at Leicester, Alexander Kraskov from University College London, Christof Koch at the Allen Institute for Brain Science, Florian Mormann at the University of Bonn and Itzhak Fried at the University of California Los Angeles.

Professor Quiroga said: “We are constantly bombarded with noisy and ambiguous sensory information and our brain is constantly making decisions based on such limited data.”

"This result supports the view that these neurons are play a key role in the formation of memory."

“We indeed see the face of a friend rather than the combination of visual features that compose the person’s face. The neurons we report in this article fire exactly to this, to the subjective perception by the subjects, not to the features of the faces they were seeing.”

“In a sense, the interpretation of this result goes way back to British Empiricism and even to Aristotle. As Aristotle put it, we create images of the external world and use these images rather than the sensory stimulus itself for our thoughts. These neurons encode exactly that.”

“This result supports the view that these neurons are play a key role in the formation of memory.”


SOURCE  University of Leicester

By 33rd Square

Embed