Hey there. I'm planning to buy a house soon and looking to source and build my first security cam system to secure it. I went from googling, to reading reddit posts (should have jumped straight there...I swear SEO has ruined practical web searching), and saw this forum mentioned as a good resource. It looks like a good knowledgeable community and I look forward to poking around.
There area few requirements I'm hoping to hit when putting together a system. I'm not sure if posting in the intro section is right, so if not I'll be sure to move it.
1) Entirely locally managed - both on the processing and storage side.
2) POE - would rather not rely on wifi or batteries, and that takes care of both.
3) 2-5 cameras - would likely just do front and back door, and possibly one camera inside to serve as a baby monitor. Maybe a couple on the sides depending on particular layout.
4) Remotely accessible via an app or web interface that plays nice with phones. Preferably through a cloudflare tunnel instead of having to run tailscale or log into the router's vpn.
and the one's I think I'm currently struggling the most with...
5) Can run from and play nice with my existing home server. I have a basic home media/home assistant server set up running ubuntu on an intel i3-13100 cpu. I really don't want to have a second device running full time when this one is always on already. Partly it's just not wanting to pay for another device, but mostly it's wanting to be efficient with electricity. I'd also like to avoid running a VM as that seems to just cause headaches, both in terms of permanently provisioning CPU cores, making use of lower level features like quicksync (hardware encoding/decoding h264 and h265), etc. If i don't have enough CPU power, I'd rather sell my CPU and upgrade to an i5/i7 rather than a whole new system. So ideally I'd find something that can run natively in linux and be spun it up as a docker container. This precludes Blue Iris, and I'm having a hard time wading through the noise, determining what's current and not, etc, when it comes to the various linux native apps (frigate, shinobi, ispy, zoneminder, bluecherry, etc). I'm hoping to find out which has the features I'm looking for, and will work best with my hardware. e.g, one that will usee quicksync (if that's even necessary/helpful).
6) Can perform some basic "AI"/"smart" type functions. I.e., can push a notification and/or flag recording for retention when a perimeter is crossed, and possibly do some more advanced things like distinguish between a person and a squirrel for instance, but not really trying to do anything like record all license plate numbers or automatically ID specific people.
To that end, one thing I'm really hoping to do here is understand how the resource allocation/workflow between the cameras and recording software works. One camera I'm looking at for instance is the IPC-T5442T-ZE. At 4MP it seems like a good balance between image quality and low light performance (I don't want a model that turns on a visible spotlight..I'd rather switch to black and white), and while the price is high, it's not prohibitively high for a couple cameras. But I just don't understand how the AI features work. I see it says that it can do things like perimeter detection, face detect, etc, and I also see that some of the software I've searched claims to do the same. So is the camera doing all the "AI" processing and just sending a signal to the NVR host with things like "intrusion detected" or "record this at higher quality." And then when software claims to do the same, is that just for cameras that don't have onboard capabilities?
I'm mostly just trying to understand how these things work before I spend a bunch of money, If anyone has any specific recommendations (e.g., the AI on hikvision cameras plays nice with zoneminder, unlike dahua, or whatever), I'd love to hear those as well though.
There area few requirements I'm hoping to hit when putting together a system. I'm not sure if posting in the intro section is right, so if not I'll be sure to move it.
1) Entirely locally managed - both on the processing and storage side.
2) POE - would rather not rely on wifi or batteries, and that takes care of both.
3) 2-5 cameras - would likely just do front and back door, and possibly one camera inside to serve as a baby monitor. Maybe a couple on the sides depending on particular layout.
4) Remotely accessible via an app or web interface that plays nice with phones. Preferably through a cloudflare tunnel instead of having to run tailscale or log into the router's vpn.
and the one's I think I'm currently struggling the most with...
5) Can run from and play nice with my existing home server. I have a basic home media/home assistant server set up running ubuntu on an intel i3-13100 cpu. I really don't want to have a second device running full time when this one is always on already. Partly it's just not wanting to pay for another device, but mostly it's wanting to be efficient with electricity. I'd also like to avoid running a VM as that seems to just cause headaches, both in terms of permanently provisioning CPU cores, making use of lower level features like quicksync (hardware encoding/decoding h264 and h265), etc. If i don't have enough CPU power, I'd rather sell my CPU and upgrade to an i5/i7 rather than a whole new system. So ideally I'd find something that can run natively in linux and be spun it up as a docker container. This precludes Blue Iris, and I'm having a hard time wading through the noise, determining what's current and not, etc, when it comes to the various linux native apps (frigate, shinobi, ispy, zoneminder, bluecherry, etc). I'm hoping to find out which has the features I'm looking for, and will work best with my hardware. e.g, one that will usee quicksync (if that's even necessary/helpful).
6) Can perform some basic "AI"/"smart" type functions. I.e., can push a notification and/or flag recording for retention when a perimeter is crossed, and possibly do some more advanced things like distinguish between a person and a squirrel for instance, but not really trying to do anything like record all license plate numbers or automatically ID specific people.
To that end, one thing I'm really hoping to do here is understand how the resource allocation/workflow between the cameras and recording software works. One camera I'm looking at for instance is the IPC-T5442T-ZE. At 4MP it seems like a good balance between image quality and low light performance (I don't want a model that turns on a visible spotlight..I'd rather switch to black and white), and while the price is high, it's not prohibitively high for a couple cameras. But I just don't understand how the AI features work. I see it says that it can do things like perimeter detection, face detect, etc, and I also see that some of the software I've searched claims to do the same. So is the camera doing all the "AI" processing and just sending a signal to the NVR host with things like "intrusion detected" or "record this at higher quality." And then when software claims to do the same, is that just for cameras that don't have onboard capabilities?
I'm mostly just trying to understand how these things work before I spend a bunch of money, If anyone has any specific recommendations (e.g., the AI on hikvision cameras plays nice with zoneminder, unlike dahua, or whatever), I'd love to hear those as well though.
