In this article, Serhii Pospielov, lead software engineer, Exadel, looks at some ways to deploy facial recognition to meet business needs and will discuss the pros and cons of each method, as well as strategies for anticipating costs.
In my previous article on facial recognition technology, we shared some use cases for this increasingly useful technology. We also briefly looked at some of the typical costs associated with implementing facial recognition tech into business. Today, we will follow that with some ways to deploy facial recognition to meet business needs and discuss the pros and cons of each method, as well as strategies for anticipating costs.
SaaS Solutions for Facial Recognition
First, let’s talk about facial recognition services and their corresponding advantages, disadvantages, and costs. When you’re first looking to implement facial recognition as a technology offering, you have options with what services you want to integrate. The first option we’ll look at is Software-as-a-Service (SaaS). Some of the big names in this area are Amazon Rekognition and Microsoft Azure, although there are other smaller solutions available as well.
Using facial recognition SaaS is the simplest way to integrate facial recognition into an existing product. In most use cases, it’s as simple as any other SaaS, where you go to their website, create an account, use their documentation to integrate, and then simply pay for the number of requests you make. If you do not use the requests, you don’t pay for them. Sounds easy and perfect, right? Well, using this method is not without its disadvantages. Let’s look at the pros and cons of this approach.
Advantages and Disadvantages of SaaS Facial Recognition Solutions
Advantages
One of the advantages of deploying facial recognition through SaaS solutions is that it is almost always the easiest to integrate. SaaS solutions have REST APIs, ample documentation, and many SDKs, so you don’t need to spend too much time integrating the solution for your needs. This appeals most to developers who don’t have to do as much heavy lifting to get the solution up and running, and to the business who can have a facial recognition solution more quickly than other approaches.
Furthermore, you don’t need to manage your own servers, which means the maintenance of your facial recognition platform storage falls on someone else’s shoulders. So, not only do you have the advantage of being able to send requests when you might have an issue, but you also don’t have to pay someone to manage your servers.
Disadvantages
Still sounding pretty good, right? Well, let’s talk about the two main disadvantages of this approach. The first is that SaaS solutions are very expensive if you have a high volume of requests. If you’re considering this approach, one of the first questions you need to ask yourself is how many requests you need to send and the costs per request. This can add up very quickly. Â
Another disadvantage is less security. There are rules around data privacy and how you handle the information of your employees, clients, and potential customers. While you may know exactly how you manage your data, the second you send it to a third party — even a reliable one — you introduce a new variable that increases risk.
SaaS certainly has its applications for facial recognition, but you need to understand how many requests you will make so you can calculate your cost beforehand. If you expect a small number of facial recognition requests or if you use a serverless backend for your business logic, then this can be a great approach.
See More: 5 Lessons To Help You Avoid AIOps Pitfalls
Self-hosted Solutions for Facial Recognition
Another option is self-hosted facial recognition services, and there are two options for where to deploy them. The first option is to deploy in the cloud, and another is to deploy on-premises. In this case, whether you buy software or use a free open source solution, where you deploy will impact cost and come with its own set of pros and cons.
Advantages and Disadvantages of Self-hosted Solutions
First, let’s look at self-hosted facial recognition services in the cloud. With all self-hosted solutions, you control where to run them, and with open-source self-hosted solutions, you control not only where to run but also how sensitive data is handled.Â
Let’s say you are using an open-source solution. If you use an open-source self-hosted facial recognition service, you control how sensitive data is handled. For the highest level of security and ensuring compliance, it’s better to choose facial recognition that’s self-hosted and open source. Another advantage of this approach is that it’s almost always less expensive than SaaS. However, the disadvantage is that because you control the software, managing it is up to you. You must rent or pay for servers and hire engineers to manage them.
With that in mind, what do the costs look like? You likely already have a DevOps engineer because every facial recognition project contains not only a facial recognition service but also custom logic that requires DevOps to deploy. If you pay for full-time DevOps support, it may be much more expensive than paying for SaaS. If you choose to deploy to the cloud rather than on-prem, you will incur costs for the servers.Â
Another approach is buying a facial recognition service license. Similar to the SaaS approach, this usually comes with a low price if you don’t have too many requests, but costs can increase quickly as your requests scale. For this reason, it’s often preferable to use an open-source solution. This means the facial recognition platform will be free, and you only need to account for the costs of your server solution.
So, when are self-hosted facial recognition solutions best for you? If you have a lot of facial recognition requests, self-hosting will likely be a more cost-effective approach.Â
Smartphone Solutions for Facial Recognition
The last option is facial recognition on smartphones. It’s a little bit different from other options because you run it on the devices of your clients. For example, the latest iPhones use facial recognition to unlock the phone. This technology may be the trickiest, and we’ll discuss its limitations.
See More: How Can Vision AI Help Airports Relaunch in a Post-Pandemic World
Advantages and Disadvantages of the Smartphone Approach
First, as a clear advantage, you don’t have to rely on the internet for this option. You also have a potentially free way of deploying facial recognition technology. So why can it be a tricky approach? The biggest reason is there are only two smartphone models that have reliable facial recognition — iPhones and Google Pixels. These newer models of phones have not only 2D cameras but also infrared cameras and more sensors so they can create a 3D map of a face. This is very reliable, but again, you are quite limited to what devices you could potentially use. So, smartphone-based facial recognition is best for you if you only need facial recognition to authorize the app, or your targets are iPhones and Google Pixel users.
Conclusion
What facial recognition solution is best for your organization? Like so many things, the answer lands firmly on “it depends.†You must consider many factors, including how many facial recognition requests you will send, how important visibility and control are for your organization, what your scaling needs are, and how your organization manages and secures its data. Once you have thought through all these considerations, choosing the right approach for your facial recognition solution should be fairly straightforward, and you’ll be on your way to putting facial recognition to work for your company.
Did you find this article helpful? Tell us what you think on LinkedInOpens a new window , TwitterOpens a new window , or FacebookOpens a new window . We’d be thrilled to hear from you.