F.A.Q.

Can 3DiVi Face SDK work offline?
Yes, Face SDK can work offline because it contains all the necessary components for the fully-fledged work that allows you to implement a variety of solutions such as cloud services, regular applications, and fully-closed systems.
Can 3DiVi Face SDK work with multiple video streams at the same time?
Yes, 3DiVi Face SDK was originally designed to simultaneously process multiple video channels in real time. It also contains built-in tools for pipelining and load balancing.
What is the maximum size of the database that 3DiVi Face SDK can work with?
3DiVi Face SDK can work with databases of any size. Technically, the software is tested on data samples of up to 300M images, but this size is not a limitation.
3DiVi Face SDK contains various recognition methods. What are they for and how to choose the best method?
Face SDK offers a set of recognition methods, allowing you to choose the suitable method based on the conditions of your case, such as the requirements for face recognition accuracy ("quality") and the speed of a single operation:
  • Method 6.6 ("Standard") has high speed and standard quality of face recognition. It is used for typical cases of real-time face recognition from streaming video data.
    Average recognition speed* is 231 ms
  • Method 7.6 ("Expert") has maximum quality but relatively low speed of recognition (lower than that of methods 6.6 and 8.6). It is suitable for face recognition in offline mode and systems with high-quality requirements for large amounts of data.
    Average recognition speed* is 1,056 ms
  • Method 8.6 ("Fast") has the best speed of recognition, but the quality of recognition is lower compared to the other methods. It is suitable for online and real-time face recognition on low-performance hardware.
    Average recognition speed* is 81 ms
* Estimated timing is given for 1 CPU core Intel Core i3, 1.0 GHz for a database of 10,000 people.
For more information on the timing characteristics of identification please check the corresponding section of the documentation
What are the hardware requirements of 3DiVi Face SDK?
Depending on the method used (see Question 4) the requirements for the optimal performance of Face SDK in the identification operations (1:N search) with 1 input data channel (camera), no more than 2 faces per frame, and a database search index up to 500 individuals are as follows:

Windows/Linux:
  • For Method 6.6 ("Standard"): Intel (R) Core (TM) i5-2400 CPU @ 3.10GHz, approximate load is 1-2 processor core(s).
  • For Method 7.6 ("Expert"): Intel (R) Core (TM) i5-2400 CPU @ 3.10GHz, approximate load is 2-3 processor cores.
  • For Method 8.6 ("Fast"): Intel (R) Core (TM) i5-2400 CPU @ 3.10GHz, approximate load is 1 processor core.
What are the requirements for size and angle of a face that ensure the normal operation of 3DiVi Face SDK?
The observed face should be located as frontally (vertically) as possible with respect to the plane of the camera lens, the distance between the pupils should be at least 30 pixels (optimal distance is 40-50 pixels) when the density of the image in the frame is at least 500 pixels/m, the permissible angles of deviation are:
  • head rotation from the frontal position (Yaw): no more than +/- 30 degrees.
  • head inclination from the frontal position (Pitch): no more than +/- 30 degrees.
  • head deviation from the frontal position in the plane of the frame (Roll): no more than +/- 90 degrees.
    How many faces in the frame can be recognized by 3DiVi Face SDK?
    3DiVi Face SDK allows recognition of an unlimited number of faces under the following conditions:
    • face requirements: faces are front-facing the camera that is located at head level (preferably), the distance between pupils is at least 30 pixels
    • hardware requirements: with a capacity of 5 people per second and a database size of up to 1 million images, the recommended CPU parameters are: 2 x86_64 cores, 3GHz each
    What set of components is required for face identification?
    3DiVi Face SDK offers two main options for the implementation of the facial recognition process:
    the use of the VideoEngine Extended component is the easiest option for typical implementation with pipelining and load balancing and, thus, it is the best solution for real-time data processing.
    the use of Face Detector + Encoder + MatcherDB (N) components is a lower-level version that is suitable for experienced developers and that allows a large number of possible customizations. In this case, the developer of an application manages the whole process of face recognition.
    A detailed description of the components is available in the documentation.
      What cameras and video formats are recommended for 3DiVi Face SDK?
      Face SDK can work with various types of cameras, including Web Cameras, Body Worn Cameras (BWC), built-in cameras (in PCs, smartphones, AR glasses, etc.), USB cameras, IP cameras, IR cameras, RGB- and RGBD sensors, and various types of Digital Video Recorders.
      Detailed information on the requirements for cameras and video formats can be found in the documentation.
        What other features does 3DiVi Face SDK have besides face recognition?
        • gender: male, female;
        • age group: kid, young, adult, senior. The age can also be estimated in the range of +/- 5 years;
        • emotions: neutral, happy, angry, surprised;
        • gaze direction (0-90 degrees, accuracy of +/- 3 degrees);
        • liveness detection that is checking if the face belongs to a real person (in comparison with the displayed photo and / or video image) or not.
        What is Liveness detection in 3DiVi Face SDK and how does it work?
        Liveness detection is a test that checks if the detected face belongs to a real person in order to prevent spoofing attacks that can be done with the help of a displayed photo or video image.
        Face SDK offers:
        • 3D Liveness Detector: a component that checks Liveness using the data from a depth sensor such as Intel RealSense, Orbbec Astra, etc.
        • 2D Liveness Detector (beta): a component that checks Liveness by analyzing face images obtained from a regular RGB camera.
        What is the difference between the 3DiVi Face SDK and 3DiVi Face Machine?
        Face SDK (Software Development Kit) is a set of libraries designed to support the development of application solutions for recognizing faces, emotions, gender, and age.
        Read more

        The Face Machine is a set of ready-made modules for face recognition in real time, consisting of a server (Face Machine Server) and a client part (Face Machine Client application). Face Machine components are technologically unified with the 3DiVi Face SDK, which means that they have a common foundation and supported integration tools.
        Read more
          What is NIST FRVT?
          NIST FRVT (National Institute of Standards and Technology Face Recognition Test) is an independent testing of biometric algorithms for manufacturers of facial recognition technologies conducted by the US National Institute of Standards and Technology under the US Department of Commerce with regularly updated results available at: https://www.nist.gov/programs-projects/face-recognition-vendor-test-frvt-ongoing

          This testing is carried out without the direct participation of vendors, so the results are considered objective, reliable and reflecting the level of development of a specific participant (vendor) in the overall rating for different data categories.
            What is the Face Recognition Terminal?
            The 3DiVi Face Recognition Terminal is a multifunctional Android device with the built-in face recognition system based on the 3DiVi biometric technologies. The Terminal is designed to work as a part of intelligent identification systems such as access control systems, time and attendance systems, payment systems with POS terminals, etc.
            Read more
              Can the Face Machine Client work with any other servers besides the Face Machine Server?
              Currently the 3DiVi Face Machine Client can work independently or with the Face Machine Server or Face Cloud.
              In the near future it is planned to realize the ability to integrate the 3DiVi Face Machine Client with external data sources and services.
                ANY QUESTIONS?
                Fill in the form and we will get in touch with you shortly