Reidentification

The Reidentification feature compares feature vectors of faces or full bodies detected on a video frame to previously stored feature vector clusters. It can create new clusters or update the stored clusters as well as save the comparison results to a Kafka topic. This way, the Reidentification feature is suitable for detecting the re-appearance of a person at a different time or location, which allows for tracking people on multiple video streams or recognizing people previously registered as staff members, returning customers, or impostors.

This feature depends on the Facial Feature Vector Extraction feature. The Reidentification feature also incorporates the properties of Feature Vector Clustering. These properties can be configured by the common configuration parameters of the two components, see Configuring Feature Vector Clustering.

This feature is provided by the UVAP Reidentifier microservice. For further information, see Reidentifier.

For a quick example setup of the Reidentification feature, see Single-Camera Reidentification Demo with Pre-clustering and Reidentification Demo with Person Names.