Optimizing motion detection performance
zpl last edited by
Hi all. I've been testing Shinobi as a docker container on an Ubuntu 16.04 NUC mini-PC system and I'm wondering if my hardware setup will be sufficient to handle streaming and motion detection for two 1080p cameras.
My hardware is an NUC mini-PC with 2-core Intel Celeron N3050 @ 1.6GHz and 4 GB of DDR3 RAM and SSD storage. I'm assuming I will want to use VA-API for hw acceleration since it has IntelHD graphics - let me know if that's not likely to work for some reason. So far I've found without hw acceleration the system grinds to a halt when I use the built-in motion detection algorithm with just one camera while recording the second.
My main questions are whether there is any performance difference (especially with regard to hw accelerated motion detection) when running Shinobi via the docker container vs. bare metal? And if bare metal is indeed better, would I be better off running Ubuntu 18.04 instead of 16.04 to get newer graphics libraries?
Thanks for Shinobi, it's looking to be a really powerful solution so far for my needs.