Support: Knowledge Base

Articles in this section

See more

Follow DF Studio People Recognition - Data Privacy

Use Case

DF Studio’s People Recognition feature dramatically improves productivity in workflows related to managing metadata about people appearing in images, by providing automated tagging and other tools. Time and effort is greatly reduced, replacing repetitive tagging labor with simpler Quality Assurance work.

Feature Overview

The DF Studio People Recognition feature provides users with a way to label people in photographs quickly with name and role information. It is powered by an AI-derived recognition service that uses a pre-trained model developed using deep learning technology to detect objects, scenes, and faces within images.1

Users may initiate the process by selecting images to process with the People Recognition feature. The selected images are sent to the recognition service which provides DF Studio with detected object, scene, and face information, including unique identifiers for the faces. DF Studio uses the face identifiers and the recognition service to associate other similar faces so that users may provide a Name and Role for a single Person, “tagging” many images at once. 

Data Privacy Statements

  • DF Studio does not train a model nor is customer content used to train a model. 2
  • Face data is stored as a per-account database of mathematical representations of recognized faces (biometric data) through the recognition service, used to associate customer-provided Name and Role information to other similar recognized faces. The face data is not directly accessible by DF Studio. It is used by the recognition service to provide similarity information between stored faces. 3
  • No images are stored in the recognition service.2
  • Face data stored for a DF Studio account is not available to other DF Studio accounts, even within the same private enterprise cloud.
  • Encryption is used for all data exchange with the service.4

Frequently Asked Questions

What data is involved?

  • Source: low-resolution proxy of user-selected image assets
  • Output: An account-scoped asset identifier, object labels, face identifiers (unique ids per face for the asset), face location coordinates within the image (“bounding box”)

How is data collected?

  • Customer selects assets within DF Studio to process using the People Recognition feature
  • Selected assets are sent to the recognition service
  • The recognition service returns asset and face identifiers along with other object and scene metadata that are stored for use in DF Studio
  • Users enter Name and Role information into DF Studio

Who owns the data?

  • Source data: images are customer-provided content, owned by customer
  • Output data: asset and face identifiers belong to DF Studio. Face coordinate data belongs to the customer.

How is this data protected?

  • Data is protected by AWS Global Network Security4

Data Retention/Deletion policy

  • All People Recognition data associated with the image (face bounding box, person/role labels) is removed when the asset is removed from the DF Studio account. Face vectors associated with a person are removed from the recognition service when there are no more images tagged with that person within the DF Studio account. All People Recognition data is removed from DF Studio and from the recognition system upon termination of the DF Studio account.
  • Data used to identify and match faces is locked in the AWS Rekognition Service Database, this data cannot be downloaded, or transferred to another system, and has no meaning without the associated AWS model.
     

References

  1. Amazon Web Services - Amazon Rekognition - What is Amazon Rekognition? (https://docs.aws.amazon.com/rekognition/latest/dg/what-is.html)
  2. Amazon Web Services - Amazon Rekognition - Data Privacy: Are image and video inputs processed by Amazon Rekognition stored, and how are they used by AWS? (https://aws.amazon.com/rekognition/faqs/#topic-15)
  3. Amazon Web Services - Amazon Rekognition - Biometric Laws (https://aws.amazon.com/rekognition/faqs/#topic-19)
  4. Amazon Web Services - Amazon Rekognition - Infrastructure Security (https://docs.aws.amazon.com/rekognition/latest/dg/infrastructure-security.html)

Related articles