Sightcorp logo

Face Tracking

Everything about Face Tracking

     

What is Face Tracking Technology?

Face Tracking Technology detects and tracks the presence of a human face in a digital video frame. This technology can be incorporated into computer and mobile applications, and can even be used in robotics.

Face Tracking Technology can be used online or offline.

Why is Face Tracking important?

Face Tracking enables the development of technologies such as face analysis and facial recognition.

When it comes to face analysis, Face Tracking makes it possible to:

  • Follow a particular face as it moves within a video stream
  • Count the number of people in a video frame or live video stream
  • Determine the direction in which a face is looking
  • Recognize facial expressions and perform sentiment analysis

 

In the context of facial recognition, Face Tracking can give greater accuracy compared to previous biometric recognition methods like iris and fingerprint recognition. It has also proven to be more secure and harder to hack, making it increasingly important for use in security systems. It plays a powerful role, especially in applications such as access control.

At Sightcorp, we use Face Tracking Technology in the following products:

 

Discover our products

What are the uses of Face Tracking?

Anonymous Face Tracking

Anonymous Face Tracking can be used by retailers to count the number of visitors and to track the movement of visitors through their stores. This data can then be used to optimize store layout, staffing, and restocking of shelves.

Digital signage providers can also use anonymous face tracking to determine how many people are viewing their displays, what their demographics are, and how much time people spend looking at their displays. This data can then be used by advertisers to optimize their campaigns.

Click on the links below to find out how you can use Sightcorp’s Face Tracking Technology in your own business:

Other uses of Face Tracking

Face Tracking for Face Recognition

The development of accurate Face Tracking Technology has led to major developments in the field of face recognition. One important reason for this is that it makes liveness detection (i.e. anti-spoofing) possible.

For example, Face Tracking is important when using some of the following techniques for liveness detection:

  • Requiring the user to blink
  • Requiring the user to turn their heads in a randomly determined direction
  • Requiring the user to smile

 

How does Face Tracking work?

A face tracking camera captures video images that are transmitted to the Face Tracking software. The software then uses AI algorithms to detect faces. Once a face has been detected, the software is able to follow that face around within the video stream and to analyze facial features and expressions in real time.

What are some of the challenges with Face Tracking?

Face Tracking can be difficult when the captured video has a low frame rate. If the frame rate is too low, the software might lose track of the face from one frame to another. As a consequence, when the person moves within the frame, the software may detect the same face as two separate people.  To overcome this problem, we recommend a minimum frame rate of 6 frames per second (fps) for accurate Face Tracking.

Extreme head poses, poor lighting conditions, and partial occlusion of the face can also affect the accuracy of Face Tracking. To help overcome the former, we have implemented deep learning methods, which we have already integrated into our DeepSight SDK.

View DeepSight Product page

Here are other articles that 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.