I’m looking into self hosted and open source nvr options and frigate looks like the right fit for me. I’m curious what hardware others are running it on and how many cameras they have. How many people are running it in home assistsnt?
I would add, Corral TPU is nearly essential. But do not forgot about video decoding acceleration. I’m running couple of 4k H.265 cameras without HW acceleration, and ffmpeg consumes good amount of CPU out of my Ryzen 3600. So pick hardware which can decode streams from your cameras.
I have it running in a Dell Optiplex that I bought used with an i7-4790 and a Coral Mini-PCIe. I had to buy an adapter for the Mini-PCIe, but that depends on your motherboard.
I only have 2 cameras – a 1080p Amcrest wifi cam, and a 4MP Amcrest wired camera. For the 4MP camera, I use a lower resolution sub-stream for motion detection, because motion detection is CPU only. Once Frigate detects motion, it then sends everything to the Coral for object detection. That’s my understanding, at least.
I found a used HP business small form factor i5 6th or 7th gen intel for very cheap, slapped a few SATA drives in the thing, and one of the M.2 Coral TPUs.
it is running 20-30% CPU load with 6 cameras on it - but they standard HD - I pumped one at 4k, and it loaded up much higher, so I scaled back to all 1080p or less. The TPU doesn’t even hit 1% from what i’ve seen. I should probably load a better TFLite model. Nothing mission critical - mostly a novelty.
I have 4 ethernet cameras feeding into Frigate inside HAOS. HAOS is running in a Proxmox VM with 4 cores, 4GB RAM, 128GB storage and an m.2 Coral TPU passed through.
The host machine is a Lenovo m910q with an i7-6700T processor that pulls about 35w, 32GB RAM and 1 TB NVMe.
Frigate is set to retain clips for 5 days, after which they are deleted. I have a Samba Backup job that runs every night and retains 10 days of backups.
With this setup, disk space never exceeds 50%, and CPU usage never exceeds 35%.
Clips are 1080p, and total storage for the entire VM is 128gb, not 1TB. Total disk usage for the HAOS VM does not exceed 64GB for clips retained for 10 days.
The hardware isn’t super important if you can get a Google Coral TPU. You can run Frigate on a Raspberry Pi that way. Without the TPU, it can be fairly CPU intensive.
I run Frigate on an old laptop and before the TPU it would run really hot. After it runs much cooler.
I’ve got 3 Wyze cam v3s running the wyze mini hacks firmware sectioned off in a VLAN that can only reach Frigate (no internet).
I have frigate running on a cheap Lenovo M900 I got on ebay for $65 that has an i7 and 8gb of memory and it actually does fairly well without the Google coral USB TPU as long as that was the only service on that system. Trying to run Frigate on my NUC with other services without a TPU caused some issues with CPU usage but with a TPU I would bet it’ll all run on the one system.
Home assistant works exceptionally well for notifying, one of my cameras I have on UDP since the signal isn’t great and get a couple artifacts that trip it up but other than that it has been much quicker to notify and more reliable than anything in the consumer market I’ve tried so far.
4 2K cameras running in an RPi 4 container. I do have a coral TPU so that helps offload a great deal of the processing. Runs great! I use the add on HA for the integration only.
You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !selfhosted@lemmy.world
A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don’t control.
Rules:
Be civil: we’re here to support and learn from one another. Insults won’t be tolerated. Flame wars are frowned upon.
No spam posting.
Posts have to be centered around self-hosting. There are other communities for discussing hardware or home computing. If it’s not obvious why your post topic revolves around selfhosting, please include details to make it clear.
Don’t duplicate the full text of your blog or github here. Just post the link for folks to click.
Submission headline should match the article title (don’t cherry-pick information from the title to fit your agenda).
No trolling.
Resources:
selfh.st Newsletter and index of selfhosted software and apps
I would add, Corral TPU is nearly essential. But do not forgot about video decoding acceleration. I’m running couple of 4k H.265 cameras without HW acceleration, and ffmpeg consumes good amount of CPU out of my Ryzen 3600. So pick hardware which can decode streams from your cameras.
I have it running in a Dell Optiplex that I bought used with an i7-4790 and a Coral Mini-PCIe. I had to buy an adapter for the Mini-PCIe, but that depends on your motherboard.
I only have 2 cameras – a 1080p Amcrest wifi cam, and a 4MP Amcrest wired camera. For the 4MP camera, I use a lower resolution sub-stream for motion detection, because motion detection is CPU only. Once Frigate detects motion, it then sends everything to the Coral for object detection. That’s my understanding, at least.
Frigate is wonderful.
I just run it on an old mini-pc that had a free pcie slot for a Google Coral chip
About 5 cameras, nothing crazy.
And yes I use it with home assistant as well
I found a used HP business small form factor i5 6th or 7th gen intel for very cheap, slapped a few SATA drives in the thing, and one of the M.2 Coral TPUs.
it is running 20-30% CPU load with 6 cameras on it - but they standard HD - I pumped one at 4k, and it loaded up much higher, so I scaled back to all 1080p or less. The TPU doesn’t even hit 1% from what i’ve seen. I should probably load a better TFLite model. Nothing mission critical - mostly a novelty.
not the most power efficient setup.
I have 4 ethernet cameras feeding into Frigate inside HAOS. HAOS is running in a Proxmox VM with 4 cores, 4GB RAM, 128GB storage and an m.2 Coral TPU passed through.
The host machine is a Lenovo m910q with an i7-6700T processor that pulls about 35w, 32GB RAM and 1 TB NVMe.
Frigate is set to retain clips for 5 days, after which they are deleted. I have a Samba Backup job that runs every night and retains 10 days of backups.
With this setup, disk space never exceeds 50%, and CPU usage never exceeds 35%.
What resolution are your clips that you have 500gb of video from 10 days.
Clips are 1080p, and total storage for the entire VM is 128gb, not 1TB. Total disk usage for the HAOS VM does not exceed 64GB for clips retained for 10 days.
Oh I must have misread. That makes much more sense.
So in your case the vm is HAOS and Frigate is running inside that?
Yes, that’s correct.
I read the title as “NVR hardware for a frigate” and was like WTF kind of self-hosting are you doing with military hardware on a warship.
Now I kind of want a warship.
What has happened in between?
That would be one hell of a project
The hardware isn’t super important if you can get a Google Coral TPU. You can run Frigate on a Raspberry Pi that way. Without the TPU, it can be fairly CPU intensive.
I run Frigate on an old laptop and before the TPU it would run really hot. After it runs much cooler.
Same, old laptop 8th gen Intel i7, TPU, Hass and multiple other dockers including wyze bridge with 6 cheap wyze cameras.
Acronyms, initialisms, abbreviations, contractions, and other phrases which expand to something larger, that I’ve seen in this thread:
[Thread #103 for this sub, first seen 2nd Sep 2023, 07:15] [FAQ] [Full list] [Contact] [Source code]
I’ve got 3 Wyze cam v3s running the wyze mini hacks firmware sectioned off in a VLAN that can only reach Frigate (no internet).
I have frigate running on a cheap Lenovo M900 I got on ebay for $65 that has an i7 and 8gb of memory and it actually does fairly well without the Google coral USB TPU as long as that was the only service on that system. Trying to run Frigate on my NUC with other services without a TPU caused some issues with CPU usage but with a TPU I would bet it’ll all run on the one system.
Home assistant works exceptionally well for notifying, one of my cameras I have on UDP since the signal isn’t great and get a couple artifacts that trip it up but other than that it has been much quicker to notify and more reliable than anything in the consumer market I’ve tried so far.
4 2K cameras running in an RPi 4 container. I do have a coral TPU so that helps offload a great deal of the processing. Runs great! I use the add on HA for the integration only.