Over the last couple of days, the Acando team has taken a deep dive into the Microsoft Face API (or cognitive services, or even “Project Oxford”) and discovered that it holds great potential for creating meaningful services.
First of all though, is understanding the API itself, and while the API is tremendously well documented, it takes a lot of practice to get it right.
First, you need to take a photo of the people you want to identify. Let’s start with this image:
The detect service will assign a temporary faceId. To make it persistent you need to create a “face list” which you can it into by calling the persongroups service. Of course, you can’t call the persongroup service without creating a “person group” first. So, you create a person group, then create a person, and then add the face from the face list to the person in the person group. Swell.
Okay, it sounds confusing, but when you got the workflow up and running, you can add more persons, more faces, and end up with a structure like this:
The next step is of course to use this information to further tag and identify faces from photographs. In our hack project we made an app where you can add people to your group and identify persons already in the group based on new photos.
Here are some screenshots: