What is the difference between an algorithm and a biometric software product?
Biometric identification technologies are becoming more and more widespread in various spheres of human life, continuing to improve and becoming more and more accurate. They are used to confirm monetary transactions, provide access to information for official use, verify identity when registering public services, record staff working hours, implement access control, etc.
Various algorithms are used to identify users, which differ in the degree of accuracy, the cost of implementation and the possibility of application in various fields.
The most widely used algorithms for working with facial biometrics (photo and video images). Their popularity was promoted by the fact that they work quickly and accurately enough, do not require contact. At the same time, their application requires taking into account the need for appropriate conditions — sufficient lighting and the operation of cameras from different angles. There is also a possibility of fraud with the use of deepfakes — image forgery by manipulating audio and video content using neural networks. However, algorithms are also created to recognize deepfakes.
There are also algorithms for working with voice, algorithms for 3D models of the head, algorithms for drawing veins, algorithms for fingerprints, iris, etc. If earlier mathematical models were used to create them, now trained neural networks are used for this.
Such modern algorithms based on artificial intelligence, which differ from their predecessors with higher accuracy, are in the leading positions of the reports of the US National Institute of Standards and Technology (NIST) , which regularly publishes data on the results of comparison and measurement of the work of various biometric algorithms.
We can't talk about the complete absence of FAR/FRR errors, taking into account the fact that biometric technologies are probabilistic, and even the most “accurate” algorithm may be ineffective in the case of incorrect installation of cameras and sensors, non-compliance with instructions for their placement, as well as interface flaws that complicate the maintenance of personnel equipment, etc. In addition, the algorithm can be “sharpened” for the equipment of a specific manufacturer and specific conditions, which affects the possibility of its application at different enterprises, even if they belong to the same industry.
The RecFaces company came to the conclusion that it makes no sense either to adapt to the ready-made “hardware” to create an identification system, or to start from “mathematics” when a ready-made “reliable” algorithm is taken, and a system is built on its basis taking into account the customer's requirements. In the first case, the solution will be a one-time solution for a specific customer or have a limited scope due to the use of equipment from a specific manufacturer. In the second case, it will take time to create a system based on the algorithm, and it is not a fact that it will be immediately “working”, and therefore it may take additional months, or maybe years, to debug it taking into account the customer's requirements.
The decision to create biometric software products is based on the results of a number of pilot projects conducted. Having collected the requests and requirements of customers, the company structured them according to various industries and needs, and compiled a kind of product map and requirements for ready-made solutions/boxes.
The RecFaces portfolio already has ready-made software products, which, of course, are based on algorithms, but unlike them they are complex solutions ready for use after delivery, and their application will not require years of debugging and refinement. The company's future plans include expanding the list of software products aimed at the corporate sector.
Developed by RecFaces, a ready-made and in-demand solution is an independent universal solution that is not tied to any of the equipment manufacturers and algorithms, which has:
- an installer won't take more than 20 minutes to complete;
- own functional and intuitive interface;
- a large layer of built-in business logic;
- a reporting system;
- notification system;
- tools for working with photos and profiles;
- automatic auto-training system;
- built-in connectors for integration with third-party solutions;
- a wide range of API functions for integration.
The effectiveness of this approach has been proved by the cooperation of RecFaces with such well-known companies as the developer of professional software Milestone XProtect for creating Milestone security video systems, manufacturers of electronic components Schneider Electric and Bosch, as well as the supplier of industrial-class video surveillance systems Pelco. And such a named approach provides the ability to quickly integrate the product with different systems that are already installed at the customer, allowing you to supplement and expand the functionality of these solutions.