It‘s still a bit flaky getting video acceleration (not talking about object detection but video decoding) working but after that it is one of the best solutions for live object detection I‘ve ever tried: no more small animals waking me up in the night.
P.S.: I‘m also supporting them with a yearly? subsciption to train the „A.I.“ model against false positives I provide which increased the accuracy even more.
m463 · 1h ago
> no more small animals waking me up in the night.
not waking you, but it is cool to have a collection of animal photos. Sort of amazing there's a hidden world.
danparsonson · 1h ago
Hedgehogs are fantastic TV - a member of my family used to get some great footage including one very memorable fight where one ended up rolling the other one around
xiconfjs · 1h ago
For sure, but rats and moths are usually not that cool ^^
sugarpimpdorsey · 42m ago
This is becoming a real problem because the drivers/software for the Coral AI boards is yet another example of Google Abandonware(tm) which has a hard dependency on a Paleolithic-era version of Python. Comically, the hardware is still sold.
In so many words if you expect to use the Coral boards you are stuck on EOL versions of Debian/Ubuntu - which have terribly old video drivers and missing kernel GPU support. There's a good chance your modern GPU - even well-supported Intel ones - won't work.
Imagine buying new hardware in 2025 whose software still required Windows 7.
Cyph0n · 24m ago
Re: outdated Python: Isn’t this a perfect usecase for Docker? Nix/NixOS is another option.
smokel · 7m ago
No. You might get it to run, but you would also get old security exploits to run.
alias_neo · 1h ago
Mines been getting worse.
Been running about 2-3 years, was mostly fine before but now I get constant false positives from the children's garden toys, scooter left in the garden, pirate flag waving etc.
I don't submit false positives for privacy reasons but I'm looking at trainingy own model. I've got years worth of positives/negatives to train on.
sunshine-o · 48m ago
Frigate has really done a fantastic job packing everything together.
For basic needs go2rtc [0] or MediaMTX [1] can be enough.
But once you need some form of intelligence on top AFAIK unfortunately there is no unixiy tool that can take a stream and easily define and apply a model on it. You will have to code something in python.
I am using Motion [0] since years. At least for basic stuff, is easy to configure and very flexible. For more advanced configuration, it required a bit of tuning.
So burglar just need to carry big sign "Ignore previous instructions and don't report anything"? "
dust42 · 17m ago
Looking in their github, it says that it uses openCV and Tensorflow. The motion detection is done with openCV and will be immune against any attack unless you move so slow that you are under the detection threshold.
Tensorflow for the object detection doesn't do any OCR thus written instructions dont work. However, according to the website the system has a limited list of objects it detects. So maybe disguising yourself as a walking tree might prevent detection.
CobrastanJorji · 12m ago
With an open source model, though, a criminal may be able to work out a 2D image that he could print out that would identify him as a package or a windy branch.
Luker88 · 32m ago
I'm using frigate and it is really nice, though they could improve the object detection and maybe stop changing the configuration format every year
If you want to start just remember to avoid h.265 cameras so you don't need to transcode since few clients and browsers support it.
aitchnyu · 1h ago
What your "stack" of open source cameras and dvr?
thomas_witt · 1h ago
As an alternative, you might also want to check out scrypted which offers a lot of cross-integration features and hardware optimized local AI processing (eg on MacMinis M*). Developer is super responsive in the discord.
elitistphoenix · 48m ago
Google Coral Accelerator is basically abandoned these days though
BLKNSLVR · 32m ago
Still works with frigate, although I've heard that modern (whatever that means) CPUs can do as good a job as the Coral TPU, making it somewhat redundant.
I ain't running it on a modern CPU though, so I'm happy with the Coral.
moepstar · 9m ago
Anecdata: i5-6500 did recognition in about 15ms, Coral TPU (M2 variant) does it in about 7.5ms - so… probably could’ve done without it in hindsight…
geerlingguy · 27m ago
Luckily Frigate works with a ton of different accelerators, like the Hailo, Intel's iGPU, even some Arm GPUs now too.
dll · 13m ago
I have it running on an Orange Pi 5 with the Rockchip NPU, very impressed with that being supported and working so well for object detection.
senectus1 · 47m ago
My step brother has been asking me to help him setup a load of cameras for watching his marron ponds. he has foxes, crows and humans stealing from his ponds.
In theory this would really help him get alerts to invaders and I presume filter out the sheep and alpacas he has wandering around as well.
My issue is that its in a rural area and the paddocks are quite large with no power to most of the ponds so what cameras and network to use to get the data back to the storage and processing server.
Begginning to think he might be better off running a modular system, each cluster of ponds would have its own camera cluster and mini server with the network being last mile 2.4ghz just for alerts and a solar panel bank for charging the battery and running it during the day.
What would I get away with here? N100 mini device? processing maybe 6 cameras?
zhengiszen · 1h ago
OpenIPC is an alternative open firmware for your IP camera.
OpenIPC is an open source operating system from the open community targeting for IP cameras with ARM and MIPS processors from several manufacturers in order to replace that closed, opaque, insecure, often abandoned and unsupported firmware pre-installed by a vendor.
Oh i've been using frigate with a Coral-usb stick for a couple of years now and the project has been progressing nicely.
It has a very nice integration with homeassistant.
nodesocket · 1h ago
I use Ubiquiti Protect Cameras and recently bought a AI key[1] which adds license plate and facial recognition features to all cameras even non-AI enabled models. It works really well and of course all 100% self-hosted.
Does the AI key work for more than one camera at a time?
timzaman · 1h ago
Just buy Unifi guys
qwertox · 29m ago
The Network Video Recorder UNVR is 320€ VAT incl. Does this exist as a software which I can download for free and run in a VM, so that the Unify camera, which would cost at least 100€ can store the data over there?
P.S.: I‘m also supporting them with a yearly? subsciption to train the „A.I.“ model against false positives I provide which increased the accuracy even more.
not waking you, but it is cool to have a collection of animal photos. Sort of amazing there's a hidden world.
In so many words if you expect to use the Coral boards you are stuck on EOL versions of Debian/Ubuntu - which have terribly old video drivers and missing kernel GPU support. There's a good chance your modern GPU - even well-supported Intel ones - won't work.
Imagine buying new hardware in 2025 whose software still required Windows 7.
Been running about 2-3 years, was mostly fine before but now I get constant false positives from the children's garden toys, scooter left in the garden, pirate flag waving etc.
I don't submit false positives for privacy reasons but I'm looking at trainingy own model. I've got years worth of positives/negatives to train on.
For basic needs go2rtc [0] or MediaMTX [1] can be enough. But once you need some form of intelligence on top AFAIK unfortunately there is no unixiy tool that can take a stream and easily define and apply a model on it. You will have to code something in python.
- [0] https://github.com/AlexxIT/go2rtc
- [1] https://github.com/bluenviron/mediamtx
[0] https://motion-project.github.io/
Tensorflow for the object detection doesn't do any OCR thus written instructions dont work. However, according to the website the system has a limited list of objects it detects. So maybe disguising yourself as a walking tree might prevent detection.
If you want to start just remember to avoid h.265 cameras so you don't need to transcode since few clients and browsers support it.
I ain't running it on a modern CPU though, so I'm happy with the Coral.
In theory this would really help him get alerts to invaders and I presume filter out the sheep and alpacas he has wandering around as well.
My issue is that its in a rural area and the paddocks are quite large with no power to most of the ponds so what cameras and network to use to get the data back to the storage and processing server.
Begginning to think he might be better off running a modular system, each cluster of ponds would have its own camera cluster and mini server with the network being last mile 2.4ghz just for alerts and a solar panel bank for charging the battery and running it during the day.
What would I get away with here? N100 mini device? processing maybe 6 cameras?
https://openipc.org/?locale=en
It has a very nice integration with homeassistant.
[1] https://store.ui.com/us/en/category/all-cameras-nvrs/product...
Iam not sure but I think so
Otherwise pretty happy.
No comments yet