The company developed around centralized processing but embraced the edge processing trend and today has solutions based on it, very useful for its customers in different vertical markets. The CEO of Neural tells us about this journey.
from the beginning worked with centralized processing:
Historically, Neural Labs always had a centralized, server-side approach to processing. Our license plate recognition systems were installed on servers and, taking the video streams from conventional cameras, we did the processing, image analysis and neural analysis to classify the characters.
Centralization brought us many benefits in countless scenarios, such as:
- Maintenance simplicity, single point of configuration (fewer points of failure)
- Possibility of using cameras from any manufacturer (even mixing them in a project
- Use of very simple and inexpensive cameras
This approach also had and has drawbacks such as:
- High bandwidth usage (although to be honest, there were few times when enough bandwidth was not available)
- Need to use large and sometimes expensive servers
- Necessary space in the data center
With this centralized architecture, Neural Labs has installed countless cities, access controls, etc. all over the world.
Neural Labs’ Path to Edge Processing:
A few years ago, we detected that the market was changing towards a more distributed approach:
- More and more sheets required edge processing using “smart cameras”
- The simple cameras that we had used centrally already had enough processing power to run a high-performance license plate recognition algorithm (no price increase)
- Several camera factories in China were starting to have embedded license plate recognition products at acceptable price and quality.
In the company we began to consider the migration to the edge with several challenges ahead.
What challenges did this paradigm shift represent for Neural Labs?
A new architecture:
The passage of our license plate recognition algorithm from an Intel architecture to an ARM architecture, the processor on which many of the SoC (system on chip) are based, was one of the most important milestones.
One of the differential aspects of Neural Labs compared to other companies has always been the ownership of our analytical video algorithms, all developed in standard c ++ and by our own R&D .
This made it much easier for us to port and optimize our engine to this new platform.
Offer a business solution:
The second challenge was the development of an application that would run inside the camera and provide business logic to license plate recognition.
Customers do not want a license plate reader, customers need that, by means of this reading and through a list of authorized vehicles, a barrier is activated, for example. Customers need solutions!
We overcome this second challenge well, we acquired the necessary technologies very quickly and we worked on the limitations of the processor.
All products at the edge needed to be seamlessly integrated into Neural Labs’ existing solution ecosystem and even hybrid solutions centralized / distributed transparently to the customer might be required.
Shared protocols (all our products use the same) and the inclusion in our centralized system of an external camera that already returns the processed data (ideally our Ghost and Edge) allows:
- The interconnection of our solutions that allows us to tackle 100% centralized, 100% distributed or hybrid projects
- That our clients join our protocols only once and then use any of our products
- Whatever the origin of a piece of information recognized by Neural Labs, it is accessible from our website, which allows total transparency about the origin of the product to customers.
· Neural Labs has solutions for any scenario, whatever the architecture required.
· Neural Labs products based on edge processing
· The first product on the edge was paradoxically the most difficult, since it is intended for very high-speed scenarios.
· Neural Ghost is a product capable of recognizing license plates up to 180 km / h, this is possible due to the characteristics of the acquisition (camera, lens, lighting) and due to the processing capacity. This product has been very well received and today it is being used in ITS, traffic sanctions, Management of low emission zones, etc.
For the second product we jumped into a radically opposite scenario: access control; For this we developed Neural Edge, a license plate recognition engine and an entire access control management application based on lists and actions.
This product works in normal security CCTV cameras that, yes, meet these requirements in terms of acquisition and processing capacity.
Once again, this newer product is having a great reception in two vertical markets:
- Companies with vertical products that only require our data (registration)
- Companies that want to use all the logic included in the product to automate access controls
The third product, the one that closes the gap between the two, is Urban Edge, a product that covers the space of the city segment.
Neural Labs, a company ready for the future
With the rise of smart cameras, Neural Labs has been able to adapt to the new reality, adding the ability to process at the edge when necessary.
Centralized and distributed systems will surely coexist for many years, and even the trend can go from one place to another as happened with computing. Be that as it may, the company is prepared to offer the best solution for each case.
Neural Labs will continue, as always, improving each and every one of the products —iterating—, improving interoperability with other systems to be faithful to the company’s DNA: always providing highly reliable, robust and open systems.