The Azure Custom Vision service enables users to easily train convolutional neural networks (CNNs) for classification and object detection. These trained models can then be leveraged by custom vision applications as could-hosted services. The custom vision service also supports the exporting of trained models, allowing them to be used on-premises and on mobile devices for near-real time vision applications.

In this developer focused session Alan will start with the basics of neural networks and explain how they can be trained to clarify images. He will them move on to more complex convolutional neural networks (CNNs) and explain the details of object detection.

The use of the Azure Custom Vision service to train CNNs will be demonstrated using both classification and object detection examples with tips on techniques for training the networks effectively. The use of these trained networks will be demonstrated by calling the cloud-hosted services, as well as using and exported network model in an on-premises application for real-time object detection using a video stream.