G’day! Long-time-no-see and this means you deserve something special. And today our special guests are… Mr. Cloud and Mr. Server. Actually in some situation they really can help and save much: time, money, life.
All the time since we started our weblog we were talking about neural networks, machine learning, metrics and evaluation methods. It’s now time to tell about our products. Moreover recently the new version of our major product FindFace Enterprise Server SDK came. But first a little bit of history, ladies and gentlemen.
Technology becomes product
When the NtechLab algorithm was created the authors wanted to test it on really large database.
At that moment the MegaFace one million photos database fitted good for testing our algorithm, and in September 2015 the NtechLab team sent its algorithm to the MegaFace Challenge.
According to the results, published in December 2015, the algorithm was acknowledged the best.
Up to that time we new that the technology can be and must be converted into products. But first we needed to show the algorithm in real work, to demonstrate future users the real case, useful, interesting and admirable. We needed an audience.
Thus we launched with our partners service called FindFace.RU. Anybody regardless of his/her faith, race, level of education and political views could download anyone’s photo in service and find the similar person (the very person on the photo as a rule) in the Russian social network VKontakte.
The service worked great and in some days became very popular. The potential clients and partners rang from all over the world. What is more, they suggested their own scenarios of future product applications: from social networks to retail and security systems.
Cloud advantages and specialties
It was obvious that people needed face recognition service for their businesses. As a result of the FindFace.RU boom we launched our first product — FindFace Cloud API, which could be very useful for those potential clients and partners who wanted not only to test the algorithm, but also to do something based on it by themselves.
The cloud platform gave the opportunity to integrate our technology into third-party solutions by REST API. Via API were sent requests for the key tasks such as verification and database faces search, and now these requests include age, gender and emotions recognition.
At the moment of launching the cloud product we certainly understood that it had its own restrictions: huge resources consumption when working with the video streams, the need to store the data in the cloud.
Moreover, it was obvious, that different business tasks needed different solutions. If you don’t need to work with video and you have only photos, then FindFace Cloud API is just the thing: it is incredibly easy to work with, it does not need to deploy the server, you do not need to install anything. The cloud solution perfectly suits for developers of web services, various mobile apps etcetera.
However, market demand is not limited to such tasks. Many companies which wish to implement the facial recognition technology in their business solutions may need a product that will:
а) work within the customer’s infrastructure
б) work with video streams (the cloud opportunities, even if such option will appear in the cloud, are limited)
Server advantages and specialties
Just after launching FindFace Cloud API the company released FindFace Enterprise Server SDK.
This principally new product could be easily integrated into infrastructure of any company, which is interested in the opportunities of their business using face recognition technology.
The server product was developed for much more complicated client systems and is useful for big businesses that pay attention to privacy and work with video streams: banks, security, retail chains, etcetera.
Nevertheless it is worth mentioning that integration of FindFace Enterprise Server SDK will not demand from any software engineers some special skills in the field of neural networks or deep machine learning.
Today FindFace Enterprise Server SDK 2.2 offers a wide range of capabilities, including face detection, verification, identification, and detection of age, gender, and emotions. It implements the cross-platform REST API, so any application can call it from the web, desktop, or mobile. Despite the Cloud product it works in client’s private infrastructure.
The new version of the server product supports video in FindFace User Interface (FFUI):
The brand new video detector can now operate for the particular area of a frame, which makes the whole work with video streams less resource-intensive. It also allows working offline via USB-dongle. In the case of cluster installations, all the server components now refer to a single local server.
You can appreciate also:
- The ability to make black and white lists, basing on current databases
- Grouping images with the same face into «person»: in order to find all the images of a certain person, joined together under unified Person ID
- The new local server with web interface allows you to control license according to its working period, the chosen parameters and functionality.
The latest version FindFace Enterprise Server SDK 2.2 is based on the newest version of NtechLab algorithm, which was recently certified by the National Institute of Standards and Technologies (NIST) of the U.S. Ministry of Commerce. According to the results, the NtechLab algorithm was acknowledged as the best in the world.
What is NIST, why the developers from all parts of the world send their algorithms there, how the testing is going on, and what the NIST certificate means if you decided to get involved into face recognition — we’ll tell about it in the next post. Stay tuned and don’t forget to leave your comments bellow.