Fregata.app: A port of Frigate to macOS

modplan

n3wb
Mar 22, 2019
14
11
US
I originally mentioned I found (suggested by ChatGPT) a port of Frigate to macOS called "Fregata" in this thread yesterday.

Since that thread was about a specific camera, I wanted to move the conversation here to limit off-topic conversation in that thread.

I'm a former Blue iris user that moved to Frigate running in Docker on my Mac Mini a few years ago, when ChatGPT suggested Fregata I immediately installed it. I could not be more happy with this app.

You get all the features of Frigate, but object detection runs super fast on the apple neural engine, other AI features run on the GPU, and video is encoded/decoded by the apple dedicated media engine hardware instead of the CPU. All things that are impossible when running Frigate on Docker, it used the CPU only.

My CPU usage on a base M4 Mac Mini dropped from 45-60% to less than 10% and object detection is lightning fast, like 2ms. On a windy day with a lot of shadows moving, the max I've seen is 412 detections per second on my Frigate stats which are pulled into home assistant, under Docker the most I ever saw was 76, the rest of the detections were dropped by Frigate, huge difference!

Everything runs cooler, the fan never comes on, power usage has dropped from ~30W to ~12W (I still have other Docker containers running too). I've got 6 Dahua cams connected.

Highly recommended if you are a Frigate & Mac user, or if you have an old Mac sitting around and want to try out Frigate.

I'm still running the Trial, but unless something breaks and they don't fix it I'll definitely be a buyer. Oh yeah, no sub-streams required, full resolution detections!

Fregata — Native NVR for Apple Silicon if you want to try it out. I can probably help with some things if you run into any trouble, I've been using Frigate for a while, just ping me.
 
  • Like
Reactions: Arjun
I originally mentioned I found (suggested by ChatGPT) a port of Frigate to macOS called "Fregata" in this thread yesterday.

Since that thread was about a specific camera, I wanted to move the conversation here to limit off-topic conversation in that thread.

I'm a former Blue iris user that moved to Frigate running in Docker on my Mac Mini a few years ago, when ChatGPT suggested Fregata I immediately installed it. I could not be more happy with this app.

You get all the features of Frigate, but object detection runs super fast on the apple neural engine, other AI features run on the GPU, and video is encoded/decoded by the apple dedicated media engine hardware instead of the CPU. All things that are impossible when running Frigate on Docker, it used the CPU only.

My CPU usage on a base M4 Mac Mini dropped from 45-60% to less than 10% and object detection is lightning fast, like 2ms. On a windy day with a lot of shadows moving, the max I've seen is 412 detections per second on my Frigate stats which are pulled into home assistant, under Docker the most I ever saw was 76, the rest of the detections were dropped by Frigate, huge difference!

Everything runs cooler, the fan never comes on, power usage has dropped from ~30W to ~12W (I still have other Docker containers running too). I've got 6 Dahua cams connected.

Highly recommended if you are a Frigate & Mac user, or if you have an old Mac sitting around and want to try out Frigate.

I'm still running the Trial, but unless something breaks and they don't fix it I'll definitely be a buyer. Oh yeah, no sub-streams required, full resolution detections!

Fregata — Native NVR for Apple Silicon if you want to try it out. I can probably help with some things if you run into any trouble, I've been using Frigate for a while, just ping me.
I'm going to have this - a Blue Iris system consumes a lot of power overtime. Also, the latest BI updates have been abysmal, resulting in crashes and re-configuring of cameras
 
I'm going to have this - a Blue Iris system consumes a lot of power overtime. Also, the latest BI updates have been abysmal, resulting in crashes and re-configuring of cameras
Yeah I used to be one of the biggest fans and proponents of blue iris. But I don't miss it even a little bit. Something was always broken. Updates would fix one thing and break something else.
 
  • Like
Reactions: Arjun
Now to find a portable hard drive for Frigata
 
I am, of course. The latest update made LPR, Facial Recognition, and Semantic Search enrichments run fully on the Mac GPU and they're now crazy fast, just like object detection already was. I've worked with the developers to fix a few bugs I've found, they typically respond and have a fix within hours. Couldn't be happier, pulled the trigger and bought a license but I'd still recommend trying the trial and reaching out to their support if you hit any issues.

Via email they told me an iOS app is likely coming soon so the notification issue for users that don't run Home Assistant will be solved. I run all my notifications through home assistant but I'm still looking forward to the native app for timeline scrubbing and camera viewing.
 
I am, of course. The latest update made LPR, Facial Recognition, and Semantic Search enrichments run fully on the Mac GPU and they're now crazy fast, just like object detection already was. I've worked with the developers to fix a few bugs I've found, they typically respond and have a fix within hours. Couldn't be happier, pulled the trigger and bought a license but I'd still recommend trying the trial and reaching out to their support if you hit any issues.

Via email they told me an iOS app is likely coming soon so the notification issue for users that don't run Home Assistant will be solved. I run all my notifications through home assistant but I'm still looking forward to the native app for timeline scrubbing and camera viewing.
wow

this is great info. thank you. the only reason ive been waiting to try this out, was that there were no native notifications or app, since i dont use Home Assistant.

How well does this run on an everyday Mac M1 (16gig ram)

Right now this runs Sighthound 24x7 for my 4 cameras *recording to a USB wd purple, and i also use this for my everyday computing, downloading, movie watching, etc.

Zero issues right now.
 
I don't have an M1, so it's hard to say, but with a base M4, all AI Enrichment stuff turned on and 8 active dahua cams it's using 5.1% CPU, and 4.9GB of RAM currently according to the stats. Turning off the enrichments would drop the RAM usage by a lot, I think it was around 2.1GB without those because it keeps those AI models in memory. It's pretty light on the CPU since it takes advantage of the Apple Media Engine for video encode/decode, Neural Engine for object detection, and GPU for the AI Enrichments which is all the heavy work.

^ thats all with no substreams, running detection on full resolution streams
 
  • Love
Reactions: flynreelow
I don't have an M1, so it's hard to say, but with a base M4, all AI Enrichment stuff turned on and 8 active dahua cams it's using 5.1% CPU, and 4.9GB of RAM currently according to the stats. Turning off the enrichments would drop the RAM usage by a lot, I think it was around 2.1GB without those because it keeps those AI models in memory. It's pretty light on the CPU since it takes advantage of the Apple Media Engine for video encode/decode, Neural Engine for object detection, and GPU for the AI Enrichments which is all the heavy work.

^ thats all with no substreams, running detection on full resolution streams
dang, 5 percent CPU with 8 cams.

damn, thats nice.
 
  • Exclamation
Reactions: Arjun
My Mac M4 Mini still runs cool to the touch, lol