Computing Your Family Resemblance

(Inside Science TV) – One of the hallmarks of being part of a biological family unit is the resemblance between each member.

Now, a computer program can tell you who you resemble most.

“We showed that our system can actually do it with very good precision,” said Afshin Dehghan, a computer engineer at the University of Central Florida in Orlando.

Computer vision researchers have developed a program that can match pictures of parents to their children.

“Instead of having to look at millions of pictures, the computer can give you the top 20, or the top 30 images,” said Enrique Ortiz, a computer scientist at UCF.

The program converts photos into a checkerboard and extracts tiny snapshots of the face. Then it compares and sorts specific facial features to make a match.

“Computers extract features and look for part of the face that is more helpful in distinguishing the related pairs from the not related pairs,” said Dehghan.

Over time, the computer learns which features are most important for a successful match.

“It will actually learn that this part of the eye is more important than maybe the other eye, or maybe the cheekbone is more important than maybe the forehead,” said Ortiz.

“The interesting aspect of this research incentive that has come up is we can actually mimic what the human brain is doing,” said Mubarak Shah, a computer engineer at UCF.

In one example, they showed results for actress Gwyneth Paltrow.

“It was able to match Gwyneth Paltrow to her mother really well,” said Ortiz.

Scientists found the strongest resemblance of family members was most commonly between mother and daughter. But it can match any two relatives. This may one day help authorities find missing children.

Try the program for free by visiting the Center for Research In Computer Vision .

Get Inside The Science:

Who’s Your Daddy? UCF Students Program Computer to Find Out

Center For Research In Computer Vision – University of Central Florida

Afshin Dehghan , University of Central Florida

Mubarak Shah , University of Central Florida