I’m working on getting rid of my Nest cameras around the house and trying to figure out my best option for an NVR. I really only have need/want for 2 cameras right now. One is the doorbell, and another is just a camera pointed at my back door/garage doors. I’ve already replaced the doorbell with a Reolink, and the other one is on the list for replacement soon.
As far as hardware I currently have the following possible landing places for Frigate(or other software):
- Lenovo ThinkCentre M910q with an i5-7500T
- This currently has a Proxmox with:
- Linux Mint VM (running Jellyfin inside here)
- Several other LXCs (adguard, docker, tailscale etc)
- This currently has a Proxmox with:
- Lincplus Lincstation N2 NAS running Truenas
- This currently has Immich running as an app
I don’t really need a ton of long term storage or video or anything. A few days would be more than enough, especially if I am able to get some smart alerts through Home Assistant as well.
I have thought about buying another mini pc of some form and dedicating it to Frigate because I understand it can be a bit resource hungry, but I don’t really know for sure if that is necessary. Another option along those lines would be to just buy a Reolink NVR.
I have also thought about moving Jellyfin to an LXC and passing the intel igpu to Jellyfin there. And then I could run Frigate inside of Docker? Do I understand that I could then share the igpu between Jellyfin and Docker/Frigate?
I know this is all kinda scattered. I feel like I almost have too many possible options and I don’t know which one is best these days.
I am currently using scrypted (https://www.scrypted.app/) on debian. I found it scrytped easier to set-up and get into than frigate.
From Frigate’s site, regarding their paid + subscription.
Frigate+ models are trained using images from actual security cameras in use by Frigate users located all over the world.
I’m hoping this is fully opt-in, but it doesn’t say.
Yes, this is fully opt-in. You have to manually choose images and submit them if you’d like to use them as training data.
Frigate isn’t a resource hog unless you enable the inference and classification stuff. If you don’t need that, and are only running 2 cameras, it should be fine.
Shinobi is another option without all the advanced junk you may not need though.
Oh awesome, that is great to know. Looks like I should do a little more Frigate research.
Definitely +1 on Frigate!!
Frigate is painful to set up. It won’t just go out and query the onvif capabilities so you have to try and figure out its RTSP url manually and ptz support is primitive. Its low resource and stable once you manage to get it to work.
Blue Iris is much easier and more capable, but uses Windows, and its a resource hog, and its paid. But if you get past that, BI is really good.
Blue Iris is by far the most capable NVR, but it’s Windows-only so you’d need a Windows or Windows Server VM. For a basic setup, Frigate is more than sufficient.
I’d say try Frigate on your ThinkCentre and see how well it runs. I wouldn’t buy new hardware prematurely.
Do I understand that I could then share the igpu between Jellyfin and Docker/Frigate?
I’m not sure about containers like LXC, but generally you need SR-IOV or GVT-g support to share a GPU across multiple VMs. I think your CPU supports GVT-g, so you should be able to find a guide on setting it up.
+1 for BlueIris. I run Linux almost everywhere else, but BlueIris was easy to set up and has been rock solid (I.e. no intervention) on a Lenovo mini pc (I think it’s @ I5 7500T) and 4 Reolink cameras for years.
Every once in a while I think to replace it, but this is one of those things that I haven’t found a compelling reason to follow through on.
