Shinobi seems to be pretty lightweight though. I’m running it in an orangepi zero recording 4 cameras 24/7 and streaming their feeds with no hardware issues from the orangepi. Would frigate be able to run on light hardware?
Object detection will be a challenge, especially for multiple cameras. It'll probably be fine if you have an Intel processor with quick sync.
I'm running Blue Iris on a Windows VM. I also have codeproject.ai on an Ubuntu VM with a Quadro P2000 for object detection (it also does Plex transcoding, the object detection doesn't stress it very much).
My previous "home server" was a raspberry pi 4 running home assistant and motioneye for 2 cameras. It was able to handle it with a reasonable amount of headroom. That being said, I couldn't imagine an SBC being able to handle object detection on top of that.
I've got frigate running on a HAOS VM as an add-on. 2 cameras both running detection with only 2 cores dedicated to frigate.
Using proxmox on an old Intel 5960x, very minimal usage I'm sure you would see reasonable results on an orangepi. I guess make a backup of your SD / NVME before wiping and testing.
Been meaning to tweak detection as it's a bit slow right now, wanting some automations built around person / presence detection utilising zone detection but it's too slow right now.