Skip to main content

ArcFaceIdentifier

Search for a face and identification in an embedding database

This module provides a pipeline element to search for a face and identification in the embedding database. This module assumes that the embeddigs have been created using the ArcFace model. By default the embedding database is a Redis database. The embedding database uses a Approximately-Nearest-Neighbor (ANN) algorithm to find the nearest embedding. The embedding database needs to be populated with embeddings beforehand, there is an example under tools/facial_recognition.

Databases (<e-database>) currently supported are:

  • Redis (redis)

Examples:

REST API:

WIP

Configuration File:

short-from : arcface-identifier:<e-database>

pipeline:
- yolov5s-face:a:0.3:0.3
- demux
- face-alignment
- arcface
- mux
- arcface-identifier:redis
sources:
- address: videos/sample.mp4
- type: file

is_redis_available​

def is_redis_available(rs)

Check if the redis database is available

Arguments:

  • rs redis.Redis - Redis database

Returns:

  • bool - True if the database is available

Examples:

rs = redis.Redis(host=HOST, port=PORT, db=0)
rs_ready = is_redis_available(rs)

create_embedding_db​

def create_embedding_db(database)

Create an embedding database

Arguments:

  • database str - The database to use

Returns:

  • storage nearpy.storage.storage.Storage - The embedding database
  • lshash nearpy.hashes.RandomBinaryProjections - The LSH algorithm

Raises:

  • RuntimeError - If the database is not available, or if it is unable to find the EmbeddingHash configuration

Examples:

embedding_storage, lshash = create_embedding_db("redis")

ArcFaceIdentifier Objects​

class ArcFaceIdentifier(Element)

Search for an identification in the embedding database and assign it to the detection.

Attributes:

  • name str - The name of the element
  • database str - The database to use
  • supported_database List[str] - At list of the supported database types

Examples:

ele = ArcFaceIdentifier("redis")

extract_params​

 | @staticmethod
 | def extract_params(model_params)

Extract the parameters for this element.

Arguments:

  • model_params List[str] - The parameters for this element.

Returns:

  • database str - The database to use

Raises:

  • ExtractParameterException - If the parameters are not valid.

Examples:

params = ArcFaceIdentifier.extract_params(["redis"])
ele = ArcFaceIdentifier(params)

process​

 | def process(meta)

Process the detections and assign the identification to the detection.

This is the main function for the element. It takes the detection and searches for the identification in the embedding database. If the identification string is found, the detection is assigned the identification. It also includes the three nearest neighbours as candidates.

Arguments:

  • meta dict - The metadata to process.

Returns:

  • meta dict - The processed metadata.

Examples:

params = ArcFaceIdentifier.extract_params(["arcface-identifier", "redis"])
ele = ArcFaceIdentifier(params)
meta = ele.process(meta)