In Azure Cognitve Services [Face API] Whats the difference Between Using SDK and Containers

Question

Clients will like a standalone solution due to poor internet around so I want to know the difference using SDK and containers for cognitive services.

Its clear about useage information being sent back for billing when using containers but whats seperates using containers and SDK?

Answer 1

Using SDK:

The SDK is an accelerator avoiding to implement the REST API calls to Microsoft backend, but it is what it is doing behind. So in that case, your payload (whether it is an image or an image Url) will be sent to the backend. In the scenario where you have a poor internet access, it can be blocking in the case of the image.

If you use an image Url, it's the backend which will get it so it's less blocking but if your access is really limited, it may be blocking.

Using containers:

In the case of containers, you have access to the same capabilities but they run locally where you deploy the container: everything is processed locally.

So if you bandwidth is limited (or don't want your images to be processed by Microsoft endpoint), it is interesting.

BUT (yes, nothing is perfect!) you still need a (smaller) internet connectivity as the container will send small payload to Azure about the billing. Those items are not sent synchronously but if you cannot send them regularly (if I remember well, it is a 15min window maximum), the container will stop working