The release version is very stable, so I am not sure the nightly is going to fix the DB issue. Did you try running the migration.sql? I think this is needed since you are coming from an old version. The update script has been rewritten and a lot of us have used it with success.
Mike, I didn't notice anything unusual, there were 2 Intel driver updates, one of them appear to be a video driver update based on the version numbering--but the running system currently shows an older
driver (32.0.101.6881) vs the updated one in system history (32.0.101.7026). It's been...
I just noticed that this also happened today with my dev/test system. This runs on the same versions of everything as my "production" system, BI/CPAI are also running on separate physical machines.
20:45:32:Response rec'd from License Plate Recognition (YOLO11 .NET) command 'alpr' (...4aa61f)...
Mike,
I ran into this today. System had been running for 13 days or so, no changes so it, it's a headless system and only thing running on it is CPAI.
----
14:06:50:Response rec'd from License Plate Recognition (YOLO11 .NET) command 'alpr' (...2068b8) ['No license plates detected'] took...
Hey Mike, I've noticed that from time to time, the plate that is returned is not the one with the highest confidence.
For example, this is the one that seems to get picked up by ALPR database, notice that it's 95% confidence and the plate is cut off.
Where as, there are other images with...
I was thinking that there might be a number of features that might be suitable to implement in NodeRed, and this might allow users to experiment with approaches and share the ones that have worked for their needs, and maybe allow others to copy and use. This would allow you to focus on adding...
Can you pair this with the state detection and then use the pattern matching to optimize the read? 3.6% of the plates I pick up are out of state, so for me, most of my plates are CA plates. However, there are a number of formats for CA. Standard plate, fancy black plate, and temp paper plate...
Same in California. The position determines if it's a O or 0, otherwise it's difficult to tell them apart.
Iowa does it correctly
And Nevada.. bless your heart (O 0 D all look alike)
After almost a week of use, here is what I can report:
The Good:
Accuracy is much better than the old version, approx 35-40% less read errors
Good performance on iGPU (Intel 11th gen) avg 77ms inference time
No errors in the log! (this wasnt the case with previous version)
Ignores symbols in...
I think Mike's new ALPR module is a big step forward, and vehicle description and speed is on the list of features to be implemented. Do you need to make some adjustments to the db schema to accommodate for the pending features?
Have you considered exposing the CRUD functionality so that...
Question about ALPR confidence level. I've noticed that the plate reads are mostly 100% confident--even at 100% confidence, I am seeing maybe a 8-10% error rate. The error rate becomes very high once it drops below 90%--Anything less than 85% is completely wrong, and it starts to hallucinate...
Mike,
I've uploaded the snapshots here:
https://filebin.net/rz06171vkmhvqyv5
All of the ones that are named bad, I could not get them to give me a bad read, even when I matched the frame exactly as I see it in BI. The one that I was able to confirm a bad read is named confirmed.
Let me know...
Hey Mike,
I'm on 1.0.3, and I am seeing some issues with high confidence but the characters are in the wrong position. This sometimes happens with the previous code, but usually by plates where there is a space between the characters, and it happened very infrequently. This seems to happen a...
I couldn't get it working on a fresh install of CPAI either without first installing the original ALPR first, then shutting down CPAI, replace appsettings.json and then installing the new v1.03, but that could have been caused by something else entirely.
If you are not getting any calls, try doing this:
Go to your BI main settings, and go to the AI tab:
Make sure your IP and port are correct. Then click on the AI Dashboard link to the right of that to confirm you can reach it.
Next, click on the 3 dots next to the "Use custom models" and...
Mike, do you have an approx time frame for the Linux release? I'll be on a 11th gen Intel with iGPU only so I wont be able to provide any feedback for CUDA, etc.
You can expand any or all of your macros and save it to a text file on BI. To do this, go to your camera's On Alert tab and create a new action. Select write-to-file as the method. This should show what BI is sending to the ALPR database.