Beginner’s Guide to Using Microsoft Cognitive Services Face API: Part 1
Microsoft Cognitive Services (MCS) offers one of the easier to use face detection api (application program interface) endpoints. No need to install complex libraries (take that Xcode!) or to mess with your path files. MCS Face API returns a detailed JSON file with information about face detection, bounding box coordinates, age, gender and more! At Kiva.org, we experimented with its use to help us answer the question if our lender community preferred pictures of borrowers who smile in their photo.
This tutorial is for a beginner to experiment with their own face detection projects. This tutorial assumes you — dear reader — have a text editor with which you can edit code, some knowledge of the python programming language (ver 3+): native python data structures and its function definition patterns, how to install and import libraries.
Also, you likely already know how to navigate your computer or execute code via “the terminal” or “shell” in a mac or linux environment. If you want to get fancy, you might want to try playing with these code snippets in a Jupyter Notebook. Knowledge of how to access an endpoint api, the JSON (Javascript Object Notation) data format will also help make this tutorial feel less strange.
The MCS Face API can detect angle of head pose, smile, facial hair, glasses, and takes a stab of identifying the subject’s emotion. While the Face API seems fairly easy to use, finding out how to get the required subscription key is less well documented. The following steps documents my process, but then I found this route to the subscription keys the next day. doh!
Step 1: Create a Microsoft Azure Account
You can use your gmail account to login. I tried creating an account with a work email, and Microsoft wanted your personal email account instead. You need to create an account, and you may have to provide credit card information even for the free tier of services.
Step 2: Get subscription key
This seems to be the least documented part of using the Face API. And you’ll have to dig pretty deep into Azure’s menu system to get the key. How to acquire a subscription key is not clearly self evident when you log in to the service.
Click the “+” next to “New” which slides opens another menu
click “AI + Cognitive Services” option which slides open yet another menu
To get to the subscription form, click “Create” under “Face API”. If you want to know the legal terms of using the API, click the Face API link. Basically, any image you pass through the MCS Face API makes their model smarter, and you better have had permission to use that image. Microsoft also reserves the right to cut you off from using their toy. If you’re satisfied with the terms of use. Click “create” to get to the form that will actually give the subscription key. You’re not there yet, but almost!
You’ll need to name your application. You’ll have to select your subscription and probably “pay-as-you-go” will be the option made available to you.
For the location, you’ll need to select “West Central US” for the free tier access to the Face API.
Pricing Tier: 20 calls /minute and 30K calls per month will probably make due for a personal project. This means you can’t analyze hundreds of photos at a time & might have to set a timer to slow your application down.
You might need to make a new resource group, and as of this writing, I’m unclear about what this option does in the context of Microsoft Azure and accessing the Face API. I just made one called “deep-learning”. You’ll need to agree to Microsoft’s Terms of Use. Click “confirm” and two subscription keys should be displayed to you, which would be a long jumble of numbers and letters.
The “key” will be needed later when calling the api. I used key 1 for my project, and I’d imagine key 2 would work as well. Copy these keys to a file, and you’ll use them in part 2. Onward…
