Using camera object detection with a modest recording PC #24
homeprimus
started this conversation in
Articles
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Recognising people and vehicles across several camera streams can be a lot to ask of a basic PC. You may only want a clip when someone enters the driveway, but the computer has to analyse the video continuously to find that moment.
If your cameras already recognise those objects and send the results over ONVIF, they can do that work instead.
I work on Guardian Eye. Our Camera events recording mode uses those camera events to start recording. The camera detects the person or object; Guardian Eye doesn't run another video detector on the PC.
So a modest PC can be enough. It still needs to handle the video streams and storage, but it doesn't also have to run the object-detection models.
For example, enable person detection on a compatible camera. In Guardian Eye, stop that camera, select Camera events, then start it again. Walk into view and check that the event appears in the history and a clip is saved.
The camera needs to expose the relevant detection events over ONVIF. A working RTSP stream alone isn't enough, and this mode won't add person detection to a camera that only reports motion.
Release notes
All reactions