Machine Learning Digs Deeper into Azure with Custom Vision, Anomaly Detector

Microsoft continues to demonstrate its commitment to practicable machine learning as a crucial component of the cloud — and the enterprise — with product announcements for Azure Cognitive Services, its cloud-based service for using machine learning to analyze vast quantities of customer data and help produce actionable results. 

Azure Custom Vision is now in general availability. This is a cognitive service that customers can use to train a machine learning algorithm to apply labels to images. The customer supplies a data set comprised of labeled images and the algorithm trains itself to this data, then tests its accuracy against the pre-labeled images. Once the algorithm has been trained, customers can test it, retrain as needed, then use Azure Custom Vision to classify images according to the customer needs. For example, if the customer has a mobile app that allows end users to snap a photo in order to identify an inventory item, then Azure Custom Vision would be able to call up inventory information about the item. Since Azure Custom Vision also runs on the "train in the cloud, run anywhere" model, it means that customers can export the models and use them on mobile devices even when there may be limited connectivity — a development that will benefit firstline workers. 

The Custom Vision Service is available as a set of native SDKs as well as through a web-based interface on the Custom Vision home page.

Microsoft is also announcing the preview of Anomaly Detector, a new service within Azure Cognitive Services. The service helps developers embed anomaly detection capabilities into apps so users can quickly identify and correct problems. For example, the service could detect an unusually high number of user login failures and use that data to investigate whether there's a problem with users accessing a service or if someone's trying to break into the system.

This is the next iteration of an Azure Machine Learning anomaly detection API Microsoft rolled out two years ago. Per Microsoft:

"Today, over 200 teams across Azure and other core Microsoft products rely on Anomaly Detector to boost the reliability of their systems by detecting irregularities in real time and accelerating troubleshooting. Through a single API, developers can easily embed anomaly detection capabilities into their applications to ensure high data accuracy, and automatically surface incidents as soon as they happen."

The company is also unveiling a new service aimed at customers who want to upgrade their data center to run virtualized applications on modern hyperconverged infrastructure. The just-announced Azure Stack HCI solutions will feature the same software-defined compute, storage, and networking technology as Azure Stack, and include simplified cloud access via the Azure hybrid services in Windows Admin Center. 

Finally, Microsoft's Azure Data Box Edge, a cloud-managed compute platform for containers at the edge, enabling customers to process data at the edge, is now available. According to Microsoft, Data Box Edge uses a physical device supplied by Microsoft to accelerate secure data transfer from a data-collecting device on the edge to Azure. The physical device resides in a customer's premises and they write data to it using the NFS and SMB protocols. Data Box Edge transfers the full data set to the cloud to perform more advanced processing or deeper analytics; this allows customers to analyze and react to IoT events with alacrity. Data Box Edge also incorporates machine learning into its platform: customers can run Machine Learning (ML) models without needing quick or consistent connectivity to the cloud.