Sightcorp logo

Facial Recognition Software

Everything about Facial Recognition Software

     

What is Facial Recognition Software?

Facial recognition software uses AI algorithms to compare facial images and determine whether they match, in other words, whether they are of the same person. While the term “facial recognition” is often used interchangeably with “facial analysis”, it is important to note that there is a big difference between these two terms. Facial recognition focuses on establishing or verifying one’s identity, while facial analysis can be used anonymously to determine demographic information, attention, and emotion of an individual or a crowd.

Face recognition is sometimes also confused with face detection. However, face detection is merely the first step in the face recognition process, and refers only to the process of locating a face within an image. Face detection on its own is completely anonymous.

Facial recognition software usually consists of various components, as it needs to perform a series of tasks. Usually, these include face detection, image processing, comparison/matching, and identifying. To find out more about the technical details of how this works, click here [link to How Facial Recognition Works article].

How does facial recognition software work?

Facial recognition software works in various ways. It can be used to compare two facial images to determine whether they are of the same person (1:1 matching) and it can also be used to determine whether a facial image matches another facial image within a database of such images (1:N matching).

Here are some examples:

1:1 matching

You can use this type of facial recognition when you want to determine whether a person is who they claim to be. If you want to verify a bank account holder, for example, you can ask the person to submit a selfie, and the facial recognition software then compares the selfie to an enrolled image of that person in your database. If the images match, verification is successful.

1:N (1:many) matching

You can use this type of facial recognition to determine whether someone is a part of a particular group. For example, if you are hosting an event with a VIP guest list, you can use a facial recognition camera at the entrance to scan attendees’ faces as they enter. This camera connects to the facial recognition software, which searches your database containing the guest list to look for a match. If there is a match, the guest will be allowed to enter via the VIP entrance.

What are the pros and cons of facial recognition software?

Face recognition software is developing at a rapid rate. The increases in the accuracy and speed of the software mean that it can be used with increased confidence in a wider variety of situations. However, as with any new technology, there are both pros and cons to consider.

Pros:

  • Improved user experience: Unlike other biometric methods such as fingerprinting, facial recognition is contactless. This makes it a quicker and more hygienic option, especially in the context of access control at airports or big events.
  • Increased security: Biometrics are harder to steal than bank cards, devices, pin codes, passwords, and so on. Additionally, facial recognition can be used as an extra authentication layer to increase security.
  • Compatibility: Facial recognition software is usually easy to integrate into existing systems and processes, and can be used across various platforms and devices.

 

Cons:

  • Legal concerns: Because the software is advancing so quickly, lawmakers are unable to keep up. Developing the relevant laws and policies takes time, therefore companies can in the meantime partially address these concerns by choosing to process data locally, by not storing any more data than necessary, and by obtaining people’s informed consent before using their images for facial recognition purposes.
  • Potential for bias: In previous studies, it has been shown that some facial recognition software is better at recognizing males than females, and that it is less accurate when matching people of color. However, by using deep learning techniques and a training dataset that is sufficiently large and varied, much of this bias can be eliminated.
  • Possibility of spoofing: When using face recognition technology to determine whether the person on the other side of a transaction really is who they say they are, you need to make sure that the software is able to identify attempts at fraud. In the past, it was possible to trick facial recognition software using a photo of someone else, a photo of a photo, or a face mask. However, with advanced liveness detection and anti-spoofing measures incorporated into the software, it is becoming much more difficult to convince it that you’re someone you’re not.


Sightcorp’s facial recognition software

Sightcorp offers state-of-the-art image-based facial recognition in the form of FaceMatch. Currently, FaceMatch is available as an SDK, which enables you to develop your own solution that integrates facial recognition with your existing systems.

Whether you want to improve your client onboarding processes in a KYC environment, introduce face recognition for streamlined access control, or introduce face recognition as an added layer of security in the context of two-factor authentication for payment verification, FaceMatch is a good choice for you!

Keep in mind that you will need C++ expertise to further develop the FaceMatch SDK into a new application or integrate it with your own application. We are currently also working on a web-based API service, which will require minimal development work to integrate. Please feel free to reach out to us if you have any questions on this.

Learn more about Sightcorp’s FaceMatch solution

Here are other articles you might find interesting:

   

Technical Specifications

The table below shows how FaceMatch SDK performs on the Labelled Faces in the Wild (LFW) dataset:

FPRTPRThreshold (Inverse of distance)
0.10.99900 ±0.002130.55448
0.010.99667 ±0.005370.59791
0.0010.99367 ±0.006050.62989

FPR = False Positive Rate
TPR = True Positive Rate

These results are an indication only and are based on the specific dataset Labelled Faces in the Wild. Customers can expect similar performance, with possible variations due to hardware and the availability of annotated data.