AWS re:Invent Insights—Amazon Rekognition

As the CEO of an AWS managed services company I go to a lot of events and conferences (talking to fellow cloud evangelists is one of the best parts of my job) and AWS re:Invent is the show of the year when it comes to sheer volume of innovation on display. I was expecting a lot of exciting new product announcements this year, and AWS did not disappoint. 

Of the 77 new launches at re:Invent, one that I see immediately driving major innovation for our clients is Amazon Rekognition. It’s a powerful AWS-based machine learning tool that can identify objects in photos and video with incredible precision. Drawing on a deep library of images, Rekognition can ID specific items, flag for inappropriate content, detect and analyze faces—the use cases are massive. You can watch the full announcement below:

Takeaways on Rekognition

AWS has now evolved Rekognition in a way that pushes it into truly awe-inspiring territory. With the new custom labels capability, users can train it to recognize absolutely anything—even very specialized and unique objects relevant to any industry. Eliminating the need for a pre-built library means we can now bring visual machine learning to any conceivable situation. And you don’t need a staff of machine learning engineers to do it.

The potential applications for this capability are immense. First up for Metal Toad will be bringing Rekognition custom labels to the manufacturing sector. Factories have a vast amount of video data—far more than any human could ever watch, let alone derive meaningful insights from. We can now teach Rekognition to ID all the objects in a manufacturing environment—machines, their components, tools, people, logos, products—unique things that would never be available in any library. Rekognition can ingest thousands of hours of video, then identify the patterns and outliers to surface insights like:

• what stations experience delays, and which are most efficient
• when and where processes are completed faster and better than usual
• the locations and times where repeated errors and issues arise
• which processes are delivering outstanding results, and which are floundering

This is the kind of business intelligence enterprises can use to radically improve their manufacturing operations—and the kind of insight that’s just not possible without the ongoing revolution in machine learning and cloud computing. And, of course, manufacturing is just one place where Rekognition can turn masses of video data into actionable intelligence. This tech is going to be a game-changer for a lot of industries, and I’m excited to help our clients put it into play.