Knowledge check

Completed

Check your knowledge

1.

How much code is required to use the Custom Vision service?

2.

How many images do you need, at a minimum, to use to train the Custom Vision service to identify an object in an image?

3.

If you're using cloud resources in the East US Region to train your Custom Vision model, which region do you need to provision resources in to run the predictions?