Introduction to Custom Vision

What is Custom Vision?

Azure’s Computer Vision services don’t allow us to perform any operations on custom data. The service was already trained on huge amounts of data. But, if there is a use-case to implement a custom image recognition service, then Azure introduces us to its Custom Vision service.

For example, if we want to build an image classifier that classifies the images into two categories, namely, a person with a face mask and a person without a face mask, we could use Custom Vision.

For this kind of implementation, we can use Azure’s Custom Vision service and push our data (images of both categories) to it. Then, we can label the images and train the model based on the images that we have provided and deploy it as a service. We can also export the model for offline usage.

Azure’s Custom Vision provides two functionalities:

• Image Classification: We can use our images, label them and train the custom model to classify the images.
• Object Detection: We can use our images, tag the objects in the images and train the custom model for object detection.

As of now, the Custom Vision services provides the above two features only. Before moving on with the chapter, let’s go over how to create a Custom Vision resource.

The steps to create a Custom Vision resource

You can follow the steps mentioned below to create a Custom Vision resource: