In this blog post you will learn:
- What is the need for facial recognition today
- Could the demo project from a Devoteam developer Bogdan Marković solve some demanding challenges related to facial recognition?
- Is there a need for specific hardware and software
- To what type of users this solution could be interesting?
What is the need for facial recognition today?
Facial recognition is one of the most intriguing biometric technologies. Due to its non-invasive approach, it is becoming more and more popular among businesses and modern cities.
Since facial recognition is primarily dependent on digital imagery it is no surprise that as cameras got better and better, facial recognition became more and more popular.
Additionally, an increase in processing power within the last 10 years allowed for use cases previously deemed too expensive or unimaginable.
“Nowadays, when we have so many options for data processing it is not a question of whether something is possible but rather which platform and architecture to use for a specific scenario.” – says Bogdan Markovic, Integration Consultant
Could the demo project from a Devoteam developer Bogdan Marković solve some demanding challenges related to facial recognition?
This project was envisioned as a demo, proving and showcasing that facial recognition can be done fairly easily and in a cost-effective manner with a combination of on-prem hardware and cloud-based infrastructure and software
Is there a need for specific hardware and software?
“As more and more business facilities, as well as public venues, are getting equipped with IP security cameras, it has come to my idea that many of those consumers could benefit from an already available infrastructure, and with an addition of relatively cost data processing on the cloud, they can implement a fully customizable and scalable facial recognition system.
Most modern IP security cameras are already equipped with SOCs that are powerful enough to do real-time facial recognition but those are just coming to the market and are relatively expensive.
Moreover, systems are vendor locked therefore not allowing usage of cameras from other vendors. Systems like those are quite locked and don’t offer a lot of customization nor do they provide data outputs that could later be used by some third-party application for analytics purposes for example.
Where the cloud comes in is primarily facial recognition itself. Instead of doing facial recognition on-prem where a predefined amount of rather specific hardware and software is needed, instead, it can be done using AWS Rekognition service which is serverless and scalable to accommodate heavier or lighter loads.
AWS Rekognition service already has APIs for real-time facial recognition sourced from AWS Video Stream. The only bottleneck in this setup is the internet connection because all camera video feeds need to be streamed to AWS where it`s ingested into AWS Video Stream and subsequently processed by AWS Rekognition.
AWS Rekognition can simultaneously do facial recognition as well as match them to known faces within predefined collections. The output of this service can be sent to AWS Kinesis Data Stream and from there processed and filtered by AWS Lambda function and eventually permanently stored in DB such as AWS DynamoDB from which it can be consumed by third-party applications.
Output made by Rekognition does not contain cropped images of recognized or unrecognized faces but rather metadata about its findings. The output contains only coordinates of the face as well as the percentage of certainty if the face is matched to one within the face collection.
Therefore, if an actual image is needed it can then be later generated combining semi-persistent data from Kinesis Video Stream and persistent data from DynamoDB.
This also allows for images to be created based upon the predefined filter only when they are needed thus saving storage space in the cloud.
As seen by the diagram, the system is quite modular and customizable to suit different scenarios and needs.
Furthermore, it takes off the burden of handling scaling since the whole setup can easily be scaled to process as many cameras feeds as necessary.” – concluded Bogdan Marković, Integration Consultant in Devoteam Serbia.
To what type of users this solution could be interesting?
All things considered, this architecture model is something that can be deemed interesting in situations where facial recognition systems need to be scalable and cost-effective.
For instance, law enforcement agencies could greatly benefit from such systems since many cities already have sufficient infrastructure (cameras and internet bandwidth).
Moving data processing to the cloud such as AWS is something to be considered in such scenarios as not only does it provide scalably and relatively cost-effective solutions for facial recognition but it also provides a high level of data security.
AWS supports more security standards and compliance certifications than any other offering, including PCI-DSS, HIPAA/HITECH, FedRAMP, GDPR, FIPS 140-2, and NIST 800-171, helping satisfy compliance requirements for virtually every regulatory agency around the globe.
Interested in finding more about Cloud technologies – visit our Cloud dedicated page.
If you want to be a part of Devoteam, follow the link and join us!
