Facial recognition terminology explained

man pointing at whiteboard

At Nirovision, we understand that the language of biometrics can be quite complex and confusing, so we have prepared a list of terms you will come across when researching and reviewing facial recognition solutions and vendors.


Actions that take place to perform one (or more) specific jobs.


The data being used to perform the job.


How and where the technology is being deployed.  


Facial recognition is the process of identifying or verifying a person in digital images or video frames, through the facial biometric pattern and data.

This process is comprised of multiple steps, each performed by a specific algorithm with distinctive characteristics, organised in what’s called a pipeline. Every vendor has their own pipeline; not every pipeline performs the same steps, nor in the same order.

Nirovision pipeline

The most common steps are:

The act of finding a face in a given image, and pinpointing where that face is located in the image.

The grouping of similar faces within a database.

Comparing a face to a specific identity. E.g. does this face match an identity in the database? This could be done for 2 purposes:

  • Verification – answering the questions, ‘Is this person who we think they are?’ e.g. checking a detected face matches an image in a database.
  • Authorisation – answering the question ‘Is this person to do what they are trying to do?’ e.g. to allow access to something, such as unlocking a door.

Identifying a characteristic of a face, such as age, gender or expression.

Not all facial recognition systems perform all of the above tasks. Classifying for example might be used within an outdoor digital billboard whereas face recognition for a workplace might focus more on matching to identify workers entering.  


  • Biometric data
    Biometric data refers to a person’s physical characteristics such their face or their fingerprint.
  • Embedding
    A series of numerical vectors that represent key features detected in the image of a face, much like a fingerprint. Embeddings can be used to organise, filter, and rank images according to visual similarity. Embeddings are generated from the image of a face, but an image of a face cannot be generated from an embedding.
  • Identity
    An identity is the representation of a person as a collection of faces and embeddings with some associated metadata. This metadata can include a name, label or other information, but may also contain none of these.
  • Track
    A record of a series of detected faces over subsequent video frames which have been determined to be the same person.
  • Retention
    The length of time in which information in a database is stored.
  • Metadata
    Metadata is the “data that provides information about other data”. In the context of a facial recognition deployment, identity metadata refers to additional information stored alongside an identity, e.g. an external ID.
  • Encryption
    Encryption is a way of scrambling data so that only authorised parties can understand the information.
  • Model
    Sometimes referred to as the identity database, the model comprises all the generated embeddings and their relationships. Facial recognition databases are usually private and managed internally by the customer or user. Users must also give consent to be added to a database. Learn more about Australian Privacy Laws.


  • IP Camera
    An internet protocol camera, or IP camera, is a type of digital video camera that receives control data and sends image data via the internet.
  • Frame Rate
    Frame rate is the speed at which IP cameras record images within a unit of time and it is usually expressed as frames per second (FPS). Each image represents a frame, so if a video is captured and played back at 20 fps, that means each second of video show 20 distinct still images.
  • Resolution
    Refers to the size of a digital image the camera produces, and is usually expressed in terms of “megapixels.”
  • Shutter speed
    In digital photography the shutter speed is the unit of measurement which determines how long the shutter remains open as the picture is taken. It’s one of three elements that affect how light or dark an image is (called exposure) which can impact the facial recognition performance.
  • Server
    A facial recognition server is a computer that monitors the streams of nominated IP cameras and process them through a facial recognition pipeline. Servers can be hosted on the cloud, or on premises. On-premise servers run software at the premises of the person or organisation using the software, rather than at a remote facility or cloud.
Deployment in Nirovision
  • GPU
    A graphics processing unit (GPU) is a specialised electronic circuit designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device. GPUs are used in embedded systems, mobile phones, personal computers, workstations, and game consoles. Some facial recognition systems require a server to run their algorithm.

Every environment’s lighting conditions are particular, and change throughout the day. These changes, despite being barely noticeable to the human eye, have a tremendous impact on the quality of faces detected. The better the data’s quality, the better the performance of the system… and it all starts with deployment considerations such as hardware quality, camera placement, and camera settings.

We hope this helps with your understanding of facial recognition. If you have any specific questions you can contact us anytime.

Nirovision newsletter

Subscribe to get fresh stories, tips, and resources delivered straight to your inbox once a month.

Thanks for signing up for newsletter.
Oops! Something went wrong while submitting the form.
By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.