/
Facial recognition using DeepFace

Facial recognition using DeepFace

This module has been introduced in version 26.2

Introduction

This document provides technical support information for the module FACIAL_RECOGNITION_DEEP_FACE, which enables:

  1. Face detection – detecting and identifying faces on images.

  2. Person management – managing identified persons and linking them to detected faces.

See Facial recognition for detailed information.

Module plugins

The module contains following module plugins:

  • VECTOR_FIELDS: Allows storing embeddings as vector fields for faces and persons

  • SIMILARITY_SEARCH: Enables searching for similar embeddings, which is used to link persons to faces

  • FACIAL_RECOGNITION_DEEP_FACE: Enables applying face detection on a record,

  • PERSONS: Enables managing persons and linking them to detected faces, see Persons for more information

Configuration

The plugin has the following additional properties beyond the general ones stored in the custom properties.

Category

Property

Description

Required

Default

Category

Property

Description

Required

Default

Ingest

Detector

The detector available in Deep Face

Yes

mtcnn

Ingest

MaxFaces

The maximum number of faces to detect on an image

No

50

Ingest

MinConfidence

Minimum confidence score, between 0 and 1, required for a detected face; faces below this value are ignored.

No

0.92

Ingest

MinFaceSize

Minimum face size in pixels (width/height); smaller faces are ignored.

No

28

Ingest

MinFaceSharpness

Minimum sharpness required for a face (value between 0 and 100); blurrier faces are ignored.

No

50.0

Ingest

MaxFaceTiltAngle

Maximum tilt angle in degrees; faces tilted beyond this angle are ignored.

No

45.0

Recognition

MinSimilarityScore

Minimum similarity score, between 0 and 1, before embeddings are considered from the same person.

No

0.85

Recognition

MinQualitativeHumanAssignedFaces

The minimum number of qualitative faces that a human must have assigned to a person before automatic assignment to other images begins.

No

1

Recognition

MinQualitativeFaceSize

The minimum width and height of a face before it is eligible to count as a qualitative face, to reach the required amount of qualitative faces, before automatic assignment to other images begins.

No

100