Facial Feature Vector Extraction

The Facial Feature Vector Extraction feature converts the facial images of a person to 1024 byte long representations named as Feature Vector.

Feature Vectors cannot be converted back to image, however they serve as basic data for reidentification and facial similarity search (also called lookalike search).

The Facial Feature Vector Extraction feature is highly dependent on UVAP's 3D Head Pose and Head Detection features as only images where the detected person is facing the camera (in good resolution) are suitable for Feature Vector creation.

This feature is provided by Multi-Graph Runner (MGR). For further information, see MGR.

Note: In an integrator application, it is recommended to aggregate multiple feature vectors of the same person in order to improve the accuracy.