What’s new in Face API?
Here in Brintia we have worked with Face API since way back when it was still a beta version, following its first-hand evolution and providing feedback to try and mitigate the shortcomings that it experienced in its early stages.
During these stages, the most notable deficiencies recorded included:
- In the past, only 1,000 people could register in each group.
- There were less opportunities in recognition reliability in circumstances such as face position (Face API had difficulty detecting faces when subjects were looking down or profiled)
Little by little Face API has evolved tremendously and has left behinds its limitations to become an innovative service for Azure, adding new methods and even improving existing ones to offer a more optimized experience for the developer. Lets have a look at some of its developments.
Face API through the years
Face API was initially released in 2016, part of Microsofts Cognitive Services and has given us access to some of the most developed face recognition algorithms and services on the market. The best thing about this service is that it continues to surprise us with new advances.
In 2017 the emotion attribute was added to the Face – Detect method which gives us information about the emotional features of the detected face. Other updates include:
- 10,000 people may now be registered in groups of people.
- Pagination support is added for the lists of groups of people (to be able to manage large groups).
- The ability to add and delete faces from different lists of faces and different people concurrently.
In 2018 they added support for massive face containers like LargeFaceList and LargePersonGroup, which support millions of faces and people. Other updates include:
- The number of possible candidates that the Face – Identify call can return from [1, 5] to [1, 100] is increased.
- The accepted image size is increased from 4 MB to 6 MB for all calls where images are accepted (such as Face – Detect, for example).
In 2019 we have started the year off with strong development from Face API, with the snapshot function being added to support data migration between Azure distance subscriptions. We can expect the same level of developments for the rest of the year so stay tuned!
It must also be mentioned that the reliability of the detection and identification has also considerably improved.
As you can see, the vast amount of new features added and the increase in Face API capabilities is remarkable. The possibility of registering millions of people in order to identify them and obtain more and better data about their faces with greater precision makes this service one of the most successful in the Cognitive Services suite. If you haven’t had the opportunity to use Face API yet, what are you waiting for? We are already exploring its advantages here in Brintia!
Interested in learning more about Face API and Cognitive Services? Sign up for our upcoming meetup in Dublin here on the 26th of February where our expert will be delving into the topic.
Check out our other blog posts on Azure Cognitive Services here.