Geek Culture
Published in

Geek Culture

How Much Does a Facial Recognition System Cost

With the help of a facial recognition system, federal agents could capture a man suspected of abuse. The tool detected him in the background of someone else’s photo at the gym, in the mirror. So, the agents were able to get to that gym, ask about the man, and eventually capture him.

This real-life story, and many others, encourage businesses to benefit from AI services and deploy facial recognition systems.

The global facial recognition market size was evaluated at $3.8 billion in 2020 and is expected to reach $8.5 billion in 2025, growing at a CAGR of 17.2%. However, the costs of facial recognition software are challenging to estimate. There are hidden expenses that companies tend to overlook and, as a result, surpass the allocated budget.

This article breaks down the factors influencing the total price and gives tips on reducing expenditure. So, how much does a facial recognition system cost?

What is a facial recognition system, and how does it work?

Facial recognition systems are considered the most reliable among biometric identification forms, such as fingerprints and iris recognition. But the technology has its challenges. The facial recognition process often occurs in an uncontrolled environment with variable lighting conditions and dynamic backgrounds. Other factors affecting recognition quality include facial expressions, a person’s age, and ethnicity. A facial recognition system is composed of five main parts:

  • Hardware: includes servers and devices responsible for capturing images
  • Connectivity technology: allows hardware devices to transmit images for further analysis either to the cloud or to other devices on premises
  • Facial recognition software: a biometric tool that extracts faces from images and matches them to the existing database of faces for identification
  • Database of faces: a collection of identities, such as an employee database or a hub for social media images
  • Client-side web/mobile app: an interface that enables users to view the results

Here is how facial recognition systems work:

After detecting a face, facial recognition software reads the facial geometry, which encompasses around 80 different elements. The key features include the distance between the eyes, eye socket depth, cheekbone shape, and jawline length. When the analysis is complete, the tool will generate a facial signature as a mathematical formula and will compare it against other faces in the repository.

According to a recent report by the National Institute of Standards and Technology (NIST), facial recognition algorithms have a favorable average error rate of 0.08%, up from 4.1% in 2014.

Facial recognition systems have many exciting applications in different industries. One simple example of this technology is photo tagging in Google Photos, where the tech giant compares faces in one image to a preexisting database of uploaded photos to identify users.

Which factors influence facial recognition systems’ price?

Hardware

Companies need to carefully consider their hardware choices to avoid getting their hopes high with a software solution that their devices can’t handle. But at the same time, firms shouldn’t overpay for computational resources that they will not use. To have a working facial recognition system, you will need to procure cameras, switches, kilometers of cables, and servers for data storage and processing (unless you’re planning to use the cloud). All these utilities will add up to your facial recognition system’s price.

Camera options

The camera type and its location depends on the desired coverage, image quality, and angle of view. For example, if the device needs to capture a person standing one meter away, a 3–8mm lens camera is recommended. Examine the location where you want to install cameras. In the case of poor lighting conditions, opt for devices with built-in features that can compensate for the lack of light and still produce images that your facial recognition software can work with. Also, some facial recognition algorithms require 3D cameras. You might consider purchasing cameras that come with pre-installed computer vision software installed and can accomplish tasks like pre-processing and face detection. This arrangement will take some load off your custom facial recognition software, thereby increasing its speed. Such cameras can cost $100 per device. Here is an example of what such gadgets can do:

Hardware configuration

There is no standard hardware configuration suitable for every facial recognition task. So, organizations need to consider their requirements carefully. If the chosen format is not powerful enough, it will cause delays and quality degradation. Let’s assume the task at hand is to identify a person posing in front of a camera. To achieve this, we need two neural network models — a face detector and face recognizer. We can process around six frames per second. With these requirements, one could make do with a low-cost graphics processing unit (GPU), and even a central processing unit (CPU) might suffice. However, if we complicate this task by tracking people’s trajectories and actions or even increasing the number of cameras, we will have to procure a more powerful and expensive GPU.

Facial recognition software

Different types of ready-made facial recognition software are available on the market. These tools vary in their features and pricing models. Here are a few examples to help you gauge how much does off-the-shelf facial recognition software cost:

You can see that different vendors offer variable pricing arrangements allowing clients to select the suitable model based on the number of transactions planned per month, the speed of image processing, and the classifiers used.

It is crucial to note that face recognition software costs cited above account for subscription fees only, which are just a part of the overall expenses. Facial recognition solution vendors allow client companies to use their APIs, but you still need to integrate them into your system. You can turn to custom software vendors to facilitate integration and build a client-side application, which allows you to reap the full benefits of the facial recognition software. For example, integrating a tool that only does face detection and identification will cost you at least $3,000, and this number will increase with the scope of the facial recognition solution.

You might also require additional features, such as enhanced security or on-premises storage for sensitive applications where you can’t risk transmitting your data to the vendor’s cloud system. Computer vision development companies can help you build additional layers of functionality to supplement ready-made facial recognition solutions. For example, as a part of a large project, we developed a microservice for anti-spoofing as middleware between the client-side application and the facial recognition software API. Such microservice costs around $10,000 — $15,000, including development, training, and deployment expenses.

So, your final facial recognition software costs will accumulate everything mentioned above. Looking solely into software vendors’ licensing prices can be misleading.

Ready-made solutions are an excellent choice when you want to launch your product fast and avoid spending money on infrastructure. If you want a solution with specialized requirements, think of a facial recognition system at a hospital where some people wear a mask and some don’t. Then it is best to invest resources into building and training a custom facial recognition tool.

Connectivity requirements

If a company operates from a remote location where different communicating objects are positioned far apart, you will need to establish a reliable communication channel. Such high-quality cables can be even more expensive than servers and cameras and will account for a large portion of the total costs of a facial recognition system.

One tip to reduce connection bandwidth requirements is to install the server responsible for video pre-processing close to the cameras. In this case, the server will analyze the stream, extract images of interest, and transmit them further to the main server, instead of streaming the whole video. This configuration can operate on 1–2 Mbit per second, even with multiple cameras, if there are no strict requirements on responsiveness and stability.

The complexity of your facial recognition solution

Logically, the more features you want to include, the higher the price is. Some off-the-shelf solutions, such as Face++, set their prices based on the classifiers used, with more complex classifiers being more expensive.

The same applies to custom facial recognition solutions. The system may include classifiers, such as face detection, face verification, face grouping, similar face search, etc. The more models you accumulate, the more it will cost. But the number of classifiers is only one of several attributes impacting complexity. Other parameters include the solution’s scalability, number of images being processed, security requirements, availability, and fault tolerance.

A simple solution that merely counts the number of faces in a picture will take a few days to build and train. The costs of developing this type of face recognition apps are around $1,000, while more complex facial recognition tools can cost tens and even hundreds of thousands.

Here is one example of a rather complex facial recognition system. An enterprise risk management company based in the US reached out to ITRex to develop a comprehensive biometric-powered cybersecurity solution to identify people based on their unique facial features. Our team opted for Microsoft Face API to deliver a solution with a diverse set of functions. The resulting system designed cyber credentials based on biometric parameters for all user groups, contained anti-spoofing measures, offered a secure data transfer channel, and had its proprietary cybersecurity protocol that can be used by third-party software. It also operated in high-load cloud environments, included microservices that enabled painless scaling and change implementation.

The costs associated with such facial recognition solution could easily surpass $500,000.

How to minimize facial recognition system costs in the long run?

System complexity directly influences the costs of facial recognition software. You can deploy a tool with limited functionality for a few thousand dollars, while a highly complex and secure solution will cost you up to $1,000,000 and even more. The location of operations also influences the total price. If you need to survey large warehouses with hundreds of cameras, hardware costs will form a large part of your expenditures. But even with complex solutions, you can plan ahead to ensure the allocated budget is wisely spent. Don’t make the mistake of opting for a cheaper tool to end up spending on security fines more than you saved on the software itself. And those fines can be rather hefty. Not long ago, Clearview AI, a prominent American facial recognition company, was charged with £17 million in penalties by the UK as it failed to obtain consent when gathering publicly available photos of British citizens and using them in training datasets.

Here are a few tips that will help you make cost-effective decisions:

  • Opt for a ready-made solution if you have standard requirements and want something to be up and running in a matter of days without spending a lot on infrastructure. However, if you have unique and complex requirements, then it is best to invest in a custom facial recognition solution.
  • Invest in your system’s security. If your favorite ready-made solution doesn’t offer any reliable options, you can hire a custom software development vendor to build an anti-spoofing middleware and additional security features. Also, make sure you are familiar with the data protection laws in the countries of your operation.
  • If you decide to provide a training dataset yourself or compose it with your vendor, work on eliminating bias. Many open-source datasets are skewed towards the white male population. So, make sure the data you use represents your target population faithfully. Don’t procure hardware with barely enough power and storage capacity to cover your current needs. Leave a margin for estimation errors and possible business expansion. Also, as your facial recognition software updates, it might become more demanding.
  • capacity to cover your current needs. Leave a margin for estimation errors and possible business expansion. Also, as your facial recognition software updates, it might become more demanding.

To sum up, the amount of money necessary for facial recognition software depends on multiple factors. And if you can take all of them into account, you will have a more realistic feel of what you can get with your budget. Don’t limit yourself to what is available on the market. If you have rather specific requirements, turn to custom software developers for tailored solutions.

Thinking about deploying a facial recognition system? Get in touch! ! ITRex computer vision experts will help you integrate with the selected vendor’s API, build client-side software, and even develop a custom facial recognition application if needed.

Originally published at https://itrexgroup.com on February 24, 2022.

--

--

A new tech publication by Start it up (https://medium.com/swlh).

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
ITRex Group

ITRex Group

Emerging Tech Development & Consulting: Artificial Intelligence. Advanced Analytics. Machine Learning. Big Data. Cloud