Microsoft has opened up the public preview of Azure Media Analytics, a set of speech and computer vision services for analysing video.
The services vary from video stabilisation and motion detection to face detection and measuring emotions, including happiness, sadness, surprise, anger, contempt, fear, disgust: you can see an example of the service analysing the emotions in a video of former Microsoft CEO Steve Ballmer on the demo page.
More organizations are using and storing video -- for training or customer engagement or surveillance -- and they need new ways of searching and analysing that mass of information, Microsoft said.
"We are introducing these services based on feedback and requirements from customers across industries, from public safety, government and surveillance to retail, education, automotive and more. With Azure Media Analytics, we aim to improve productivity by cutting down the manual effort required to manage, process and review high volumes of video," said Milan Gada, principal program manager, Azure Media Services.
Microsoft said the tools could, for example, be used to analyse customer service calls to extract data about common complaints, or to review large volumes of user-generated content automatically to spot any inappropriate content.
The technology could also be handy for surveillance, as Microsoft notes: "With the growth of IP cameras, there is an explosion of surveillance videos. Manually reviewing surveillance video is time intensive and prone to human error."
As part of this announcement, Azure Media Indexer (which became available in September 2014) and Microsoft Hyperlapse for Azure Media Services will become part of Azure Media Analytics. Microsoft said all the services in Azure Media Analytics (with the exception of Indexer) are available at no cost for a limited time.
Customers who use Azure Media Analytics will have access to this set of services:
Indexer: Azure Media Indexer 2 Preview now supports English, Spanish, French, German, Italian, Chinese, Portuguese and Arabic, with support for additional languages expected in the coming months.
Hyperlapse: Combines video stabilization and time lapsing to create quick videos from long form content. As well as creating time lapses it can also be used to create stable videos from shaky videos captured via cell phones and camcorders.
Motion detection: This service allows users to detect motion in a video with stationary backgrounds. This is ideal for customers who want to check for false positives on motion events detected by surveillance cameras on the surveillance video feeds.
Face detection and emotions - Using this service, you can detect people's faces and their emotions: uses include aggregating and analyzing reactions of people attending an event.
Video summarization - Video summarization automatically selects interesting snippets from the source video.
Video Optical Character Recognition - This services can be used to extract text that is displayed in the videos, useful in building a smart search engine for videos which has a lot of text.
Content moderation - This service is designed to cut down the manual effort involved in reviewing content for offensive, illicit and inappropriate material.