Documentation for Face Intellect 8.0-8.1. Documentation for other versions of Intellect is available too.
Previous page Next page
The face recognition server stores three types of face images:
- Reference faces are face images compared with images of faces fixed on a video image.
Reference faces database is filled manually in the Face recognition and search interface object (see Adding images to the reference face database). - Captured faces are face images recorded on a video image.
Photos of captured faces can be viewed in the Face recognition and search interface object (see Configuring the Face recognition and search interface object). - Recognized faces are captured faces who have shown a sufficient degree of similarity to the reference faces.
Photos of recognized faces can be viewed in the Face recognition and search interface object (see Configuring the Face recognition and search interface object).
In addition, during the recognition process, the following data is created and stored:
- Vector is the mathematical representation of the face, created when it is recognized in the frame of the video image.
- Metadata is the data that enables linking the reference, captured and recognized faces to each other, for example, their identifiers.
Attention!
With an increase in the number of faces in the database, the statistical error increases: the more faces in the database, the more often similar faces will be recognized. Accordingly, the degree of similarity when comparing the reference face with the captured face will decrease.
This statistical error is relevant if:
- The Requirements for images uploaded to the reference face database and requirements for installing and configuring video cameras for the corresponding software module are met (see Configuring the program modules).
- The database contains over a million faces.
An example of the results of calculating the error:
- VA module, mugshot dataset (good quality photo), 12 million faces in database, and false matching probability is 0.003%. With these initial data, the researchers obtained an identification error of 0.76%.
- VisionLabs module, the initial data are the same. The identification error is 0.81%.
The size of the face database can be adjusted by Configuring the face images, vectors and metadata storage depth.
Overview
Content Tools