Video: How to Use AI for Live Stream Analysis

Learn more about video AI at Streaming Media's next event.

Read the complete transcript of this clip:

Jun Heider: Live stream analysis is pretty sweet. This image you see here is Valossa at NAB. I went to their booth and they had this live stream playing and during the playback, it was putting bounding boxes around logos, it was giving me an indication of what time of day it thought it was, and just telling me some other things about the objects it saw on the video.

Not all of these services do live, yet. I heard there's a panel earlier, and there's a Microsoft guy in there, he mentioned something, he alluded to live next year some time, but there's no details; I'm sure we'll find out soon.

But who's doing it right now? Valossa is doing it right now. AWS actually has a really, really compelling use case when it comes to live as well. They do kinesis video streams, and what they allow you to do is you stream your video, and at certain points they send the frames down, they can do facial detection right out of the box.

At NAB, they have this camera sitting on top of a TV in their booth and as people are walking by, and they had this application that was showing how much traffic was walking in front of that camera throughout the course of the conference. That was pretty sweet. And it was using facial detection to figure that out. And that's one that you can use.

You can use that SD CAdES and not just do facial detection, but have it do other things, such as object detection or speech-to-text. Hive has a live as well. They have an enterprise solution called Hive Media.

So if you're an enterprise and you want to so some of that live processing, you may want to check Hive out. And then Veritone, they can do live as well. And they actually have a really cool blog post on AWS's developer area where they're talking about how they leveraged some things like kinesis video streams and Fargate to like do some live detection.