Stream live video with metadata on local network to webapp

Iv'e seen similar questions being asked but not for this specific use case.

I have a system with four edge devices. Each is connected to a camera and reads live video from it. It will then perform some deep nn analysis on each frame and will create a data structure for each frame it processes (json right now). I want each edge device to stream the video from the cameras with with the data for each frame to a centralized server on the same network where the edge devices exist. The videos should be displayed on a web browser displaying the data for each frame received.

We have some experience in the past with using a pub-sub type messaging for this. Each message consisted the frame and data. It was usually very network and cpu intensive and very hard to scale. Its was probably not the right way to do this.

Which technology would be best to use for this case?