Early adopters and potential adopters of facial recognition technology have been pressing hard for improved performance. Animetrics and others have responded by coming up with automated ways to improve the quality of the data processed through to the template generation software for matching. The press release does a very good job of describing Animetrics approach.
Animetrics Unveils ID-Ready, Cloud-Based Facial Biometric System (Animetrics Press Release)
The service takes a grainy, partial view, angulated 2D facial image, applies 2D-to-3D algorithms and corrects the pose of the face, and makes it ID-Ready for most any facial recognition system.Read further and you'll discover that the innovation doesn't stop with the technology. The sales/distribution model is noteworthy, as well.
“ID Ready essentially takes a bad image and makes a mugshot out of it,” said Paul Schuepp, chief executive officer of Animetrics.
Most facial recognition systems require photos be a frontal view of a face in order to make a positive match. However, most photos studied by law enforcement are of faces that are rotated, “off pose” and are captured by low resolution video security cameras or long distance telephoto surveillance cameras.
“This type of uncontrolled imagery renders face recognition systems impractical because of the poor matching results, if results occur at all,” says Schuepp.
Here’s how the system works: law enforcement personnel upload a 2D photo to Animetrics servers at id.ready.animetrics.com and the ID Ready system applies facial feature point detection (eyes, nose tip, mouth, etc.) to accurately find the face and specify the parts Fine-tuning is possible by the user positioning three red crosshairs over both eyes and tip of the nose.
From there a 3D model is created and a new 2D resultant image that is pose-corrected to zero for facial pitch, yaw, and roll along the x, y and z axis.
