Page 12 - MSDN Magazine, December 15, 2017

Figure 3 Realtime Crowd Insights Solution with Cognitive Services
ees via Web sites, applications, text/ SMS, Skype and more.
Tools like Visual Studio Tools for AI, Azure Machine Learning Studio and Azure Machine Learning Workbench provide a great starting point to get started building inno- vative, intelligent AI applications.
Let’s get started with Microsoft AI by using the various services to build an AI application that lever- ages the intelligent cloud and can be deployed to the intelligent edge. I’ll start with Cognitive Services, then move on to building custom mod- els with Azure Machine Learning. I’ll finish with a dive into the Bot Framework and show how you can turn any bot into an intelligent bot powered by Microsoft AI.
upload to the Custom Vision service, as depicted in Figure 2. This state-of-the-art system allows you to develop a highly performing computer vision model in just minutes.
Custom Machine Learning and Deep Learning Models:
As you work on various use cases, data scientists in your organiza- tion might need to develop and customize deep learning models, using various deep learning toolkits. The Microsoft AI platform provides an open and flexible environment for that deep learning. Azure Machine Learning empowers data scientists to build, devel- op and manage models at scale, while data stores like CosmosDB, SQL DB, SQL Data Warehouse (DW) and Azure Data Lake (ADL) provide access to the structured and unstructured data that inform your ML and deep learning models.
With Azure Machine Learning, you can easily train your mod- els in Spark, run them on Azure Deep Learning Virtual Machines (DLVM), or process them on a managed GPU cluster with Batch AI, and more. Azure Machine
Learning experimentation and
model services boost productivity
by helping you keep track of your
projects, enabling you to train on
both local and remote compute
infrastructures, create contain-
ers for model deployment, and
manage and monitor the behavior
of models.
Cognitive Services
You can add AI capabilities to any .NET app you’re developing using Cognitive Services. Let’s get started with the Intelligent Kiosk sample app found on the .NET Machine Learning and AI Web site at bit.ly/2yNtRpF. You can access the source code from bit.ly/2zysSXJ.
The Intelligent Kiosk sample app shows how to use Cognitive Services in different scenarios, including:
• Customizing product recommendations based on detected gender and age of visitors.
• Building a real-time AI pipeline to analyze visitor demographics for retail.
• Performing automatic photo capture and face identification. Let’s consider the scenario of building a real-time AI pipeline to analyze the demographics of people visiting a retail store called Realtime Crowd Insights. In this scenario, the Realtime Crowd Insights app continuously analyzes key frames from the live video
Bots provide exciting new ways to engage with customers and em- ployees, helping them complete tasks. The Bot Framework pro- vides a rich set of capabilities for conversational AI, so you can devel- op powerful new bots that interact with your customers and employ-
8 msdn magazine
Figure 4 Creating a Cognitive Service and Obtaining the API Key Using the Azure Portal Artificial Intelligence