Documentation for Axxon One 2.0. Documentation for other versions of Axxon One is available too.

Previous page Face detector TV  Hardware requirements for the Face detector TV and its sub-detectors Next page

Description

Attention!

Starting with DetectorPack 3.11.1, the Face detector TV is no longer supported. We replaced it with the Face detector VA. You cannot use the Face detector TV and the Face detector VA at the same time (see Manual for switching from the Face detector TV to Face detector VA).

The Face detector TV generates an event every time a face is captured in the frame.

The Axxon One database stores all faces in binary form:

  1. All captured faces are stored as biometric vectors (see Face search) in the t_face_vector table, and their corresponding capture events are stored in the t_json_event table.

    Attention!

    The default storage depth of face vectors is 24 hours. To specify the storage depth of face vectors, you must add the NGP_FACE_VECTORS_IN_RAM_DEPTH_HOURS system variable (see Appendix 9. Creating system variable). The values of the variable are specified in hours.

    For example, if you set NGP_FACE_VECTORS_IN_RAM_DEPTH_HOURS = 1, face vectors are deleted from RAM in one hour. If the value of the NGP_FACE_VECTORS_IN_RAM_DEPTH_HOURS variable isn’t set, the storage depth of face vectors is 24 hours.

  2. All reference faces from the lists (see Faces) are stored in the t_face_listed table

Attention!

When the number of faces in the database increases, the statistical error increases: the more faces in the database, the more often similar faces are recognized when searching in the archive. Accordingly, the degree of similarity when comparing the reference face with the captured face decreases.

This error is relevant if:

  1. The Video stream and scene requirements for the Face detector TV and its sub-detectors are met.
  2. The database contains over a million faces.

Example of error calculation results:

The mugshot dataset (good quality photo), 12 million faces in the database, and probability of false matching is 0.003%. With these initial data, the researchers obtained an identification error of 0.76%.

LicensingTo operate the detector, contact an AxxonSoft manager and get a license file (see Types of licenses)
Add-ons

For the Face detector TV to work, you must install TV Face Recognition Pack add-on (see Installing DetectorPack add-ons).

Attention!

  • You can install only one Face Recognition Pack: VA Face Recognition Pack, VL Face Recognition Pack, or TV Face Recognition Pack. You cannot install several packages in the same system.
  • TV Face Recognition Pack is guaranteed to work on Linux Debian 11 and Ubuntu 20. TV Face Recognition Packs isn't guaranteed to work on Linux Debian 9/10, Ubuntu 18 and Ubuntu 19 (see Operating system requirements).


  • No labels