Local license plate detection without containers?

Is it possible to get cameras to detect license plates (and read them), with just using my Generic x86-64 system that is running Home Assistant only, with Coral TPU usb attached.
HAOS is installed directly to my NUC and it is not running containers.
I found this Guide: Local License Plate Recognition With Home Assistant, CodeProject.AI, and Frigate guide, but it uses container to install CodeProject.AI Server if i understand correctly.

Have a look at Platerecognizer. There’s a custom component fro HA. Been using it for years.
There is a free tier

1 Like

Isnt that cloud based? i was looking for local solution if possible.

Have a look at “openalpr” enough to find…

I did and found “nothing”, it seems openalpr local is removed from HA and all i can find was issues that it is not working.
From HA china website i found openalpr local install instructions. And GitHub - samtwilliams/ha-openalpr: OpenALPR Number Plate Recognition for Home Assistant from 6 years ago.
Is it still worth installing addon that dates 6 years ago? Will it be more troubles than solutions?

It seems this is one of your options yes…

Another option

Edit: partly cloud based

Have you found something? I run HAOS supervised also - NO containers. Addons are basically Docker containers but so far nobody feels the need (or like me they have the motivation but not the necessary knowledge or skill) to write an addon.

Depending if you want local only or cloud based - as previously mentioned have a look at plate-recognizer if cloud based in an option for you

1 Like

Hi,

the new frigate 0.16.0 Beta 1 has a licence plate build in now:

So perhaps, this could be an alternative.
Ronny

1 Like

Thanks. I must’ve tried that a few years back and lost interest - I went to get an api and I already have an account. It was no longer in my configuration.yaml so I may have found it too difficult; not so today as I have all that HA experience under my belt LOL. I have it set up to read plates when Frigate detects cars on my garage camera; so far the images have been too low quality though. Might be time to upgrade to a 12mp??

1 Like

Thanks for the advice.I have the alpr running for now per BambamNZ - I didn’t see your post. I can wait for the stable release of 16 and can get face recognition then. I am so appreciative of people like blakeblackshear and crzynik who not only write this code but also help less advanced people like me. And these forums and people like you and BambamNZ who take the time to offer advice and education

1 Like

As I mentioned earlier, I have plate recognizer working - although not well - now. I have it in an automation that triggers when Frigate integration detects “car” on an Amcrest 5mp camera. The video from Frigate are clear enough but every time the action: image_processing.scan
target:
entity_id: image_processing.platerecognizer_garage_2
runs it takes a snapshot of a blurry image and returns “null”. It is my driveway so the fastest I am going is maybe 5 mph.
Anyone know if the licesnse detection in Frigate 0.16 Beta is better? I had ChatGpt
TRY to help me set up an automation to take 5 snapshots and analyze them but place recognizer won’t allow quick succession submissions (according to ChatGPT) and spacing them 1 second apart took about 12 seconds to analyze all 5. Hopefully (if 0.16 does this locally) it will have the ability to analyze multiples faster and more reliably

All details on frigate are in the linked in the extra setup page:

If you have the 0.16 beta you can try it, all is done locally. I have no cameras for this use case, but enabled it, the cars are to fast to catch here :slight_smile:

1 Like

I installed 16.0 beta. Facial rec seems pretty accurate so far - still training, but it’s pretty much just my mug around here LOL. Lpr is buggy - maybe my cams aren’t fast enough too. I upped my detect fps a little on 3 cams but didn’t have much effect. Using HAOS so can’t get hw accel to work with Intel iGPU. Not sure that would help anyway. Planning on getting a Frigate+ subscription; I hear it will help Lpr. Worth a shot and it helps support the awesome developers. Thanks for the direct.