Face recognition engines

Face recognition engines analyze human faces in media assets and score them against the similarity to known faces. They expand on the capabilities of face detection engines by identifying the individual whose face was detected in addition to specifying where in the image the face is located.

Training and libraries

Since all face recognition engines identify entities, they are all required to be trainable via libraries.

Engine output

Face recognition engine output should be stored as objects in the time-based series array in AION. Each object should be of type face, and because each recognized face always maps back to an entity in a library, each object should always include the entityId of that original entity in its output, along with the libraryId where it can be found.

Example

Here is an example of the simplest type of face recognition output:

{
  "schemaId": "https://docs.veritone.com/schemas/vtn-standard/master.json",
  "validationContracts": [
    "face"
  ],
  "series": [
    {
      "startTimeMs": 9800,
      "stopTimeMs": 17200,
      "object": {
        "type": "face",
        "entityId": "11a14999-0531-4d3e-9a44-68cdd4f93659",
        "libraryId": "13e6f4a3-0d5c-4e11-9a30-913e981cb9ad",
        "confidence": 0.9772560119628906,
        "boundingPoly": [
          {
            "x": 0.1,
            "y": 0.1
          },
          {
            "x": 0.1,
            "y": 0.5
          },
          {
            "x": 0.5,
            "y": 0.5
          },
          {
            "x": 0.5,
            "y": 0.1
          }
        ]
      }
    }
  ]
}

• Learn about AION.
• See the cognitive engine classes.
• See an overview of cognitive technology.
• View engines for your media type.

• Using aiWARE
• Use cognitive engines
• Choose an engine
• Capability
• Biometrics
• Face recognition