What is Face Recognition Software?

Face recognition software is a computer application which identifies a person’s face appearing in a photo or in a video. Face recognition software is widely spread and can be used for many different applications, such as video surveillance for airports and casinos, demographic (age, gender and ethnicity) analysis for shopping malls, law enforcement agencies, attendance systems for school, logical access control to a bank account or physical access control to a data centre, and many more. For more information about the different applications please visit our products and solutions page.

How Face Recognition works?

Face recognition is done by comparing face representations rather than comparing actual faces. This means that a facial recognition algorithm goes through a face and extracts the features of the face. This process is often called template/vector generation or facial feature extraction. At the end of that process, we are left with 5KB of a text string which represents a 2, 5 or 10 MB photo. So when the matching process takes place, it matches a text string against a data base of text strings, which is obviously faster and much more efficient than comparing photos.

What are the Facial Recognition Steps?

The first step for using facial recognition software is to enroll (register) people into the software watch list (data base). Enrollment is done by using high quality reference photos – these can be sharp and big passport photos like. The second step is to feed the software with high quality videos or photos where the faces are fully visible. The next step is the matching process where the software compares the faces’ representation in the video or in the photo against the ones in the watch list. The matching can be done either as an identification process (1:N), where comparison is run against the whole data base, or as a verification process (1:1) where the software verifies the identity of a known person against its presumed photo in the data base.

How Accurate is Facial Recognition Software?

Our facial recognition software is highly accurate! There are several techniques to measure face recognition accuracy, but generally speaking our software reaches 99% accuracy under a control environment. Although real life conditions might be dynamic and can’t always be controlled, it is quite possible to pre-configure all the necessary requirements and to achieve close to optimal conditions.

Requirements for Achieving High Accuracy

Face recognition is not magic. Enrolling low quality photos of faces into the data base, and feeding the software with videos where faces are barely visible – are a recipe for disappointment. There’s an old saying in the software world – “garbage in – garbage out” and face recognition is no different. In order to achieve high accuracy, first you need to make sure the faces in the photos you enroll into the software’s data base are of high quality. Secondly, you need to make sure your camera is located at right place for capturing fully visible sharp faces, and then you need to make sure lighting conditions at the camera’s range are fine. For more detailed requirements of achieving high accuracy please refer to our photo guide.