Blue Iris and CodeProject.AI ALPR
- Thread starter MikeLud1
- Start date
Quick update I was barking up the wrong tree with the custom models. It looks like blue Iris isnt connecting to the CPAI server even though its online and works when tested with a test image
Ive seen other people complaining about this but cant find a solution. Blue iris is configured to not use CPAI as a service so dont really know where to go from here
IIRC I had this issue and ended up having to turn off Object Detection under Motion Sensor in Trigger tab. Maybe that's on
Thanks the for the suggestion. It was turned on so now turned off but still getting the error message that "AI cant connect to localhost:32168" This is so frustrating as I think I am really close with getting my GPU etc working
Ah, ok, thank you!
I was already trying something similar, but I was overengineering it a bit. I added the plates to the "myplates" list in the general AI settings and then put "myplates" in the "to cancel" box, but that didn’t work. I’ll try adding the plates directly instead.
Should I remove "DoNotCancel," or should it remain alongside the plates?
I was already trying something similar, but I was overengineering it a bit. I added the plates to the "myplates" list in the general AI settings and then put "myplates" in the "to cancel" box, but that didn’t work. I’ll try adding the plates directly instead.
Should I remove "DoNotCancel," or should it remain alongside the plates?
Thanks to all!
As a new subscriber to this list, I read over everything about doing ALPR and now have a setup with BI, CPAI (on a different host using only CPU, though I have a Coral TPU for it as well), and an Empire Tech camera(the famous Z12 recommended here and available from Amazon). I followed all the guidelines and other great ideas I've read here, but to avoid being intrusive in my neighborhood, I put the camera inside my garage, with careful pointing out through the typical small garage door windows to see the street outside. For daytime use, it works beautifully, but I wasn't able to really test night time views with the configuration. Obviously, the night was too dark even with a nice street lamp nearby, and I couldn't shoot the IR illuminator on the camera through the garage door. BUT, I did set up a night profile using the advice found herein, and I tested by opening the garage door temporarily to allow the camera's inbuilt illuminators to do their work. After a car passed by, I ran and checked the footage and found a near perfect license plate exposure on the first attempt!!
I obviously have some more work to do to get a more permanent camera placement for day/night use, but thanks to all the folks in these forums, and especially the superstar ALPR geniuses like MikeLud1, this looks super successful so far! I am continuing to do various tuning of zones and such to perfect the daytime collects, and will try to tackle Node-Red integration next.
THANK YOU!
I'll also add that CPAI really made this whole thing possible for me, and for many others, I suspect. I had tried in the past to deal with all the messy software installs and versioning hassles to get some AI guts working for this kind of task, whether on a Raspberry Pi or a Jetson Nano. But it proved to be a nightmare, and all the good tutorials proved to be almost useless the next time a software update came out for one of the MANY software packages involved, as the new updates invariably broke something that I was not qualifed to troubleshoot.
So, the Code Project AI folks are heroes in my book!!!
As a new subscriber to this list, I read over everything about doing ALPR and now have a setup with BI, CPAI (on a different host using only CPU, though I have a Coral TPU for it as well), and an Empire Tech camera(the famous Z12 recommended here and available from Amazon). I followed all the guidelines and other great ideas I've read here, but to avoid being intrusive in my neighborhood, I put the camera inside my garage, with careful pointing out through the typical small garage door windows to see the street outside. For daytime use, it works beautifully, but I wasn't able to really test night time views with the configuration. Obviously, the night was too dark even with a nice street lamp nearby, and I couldn't shoot the IR illuminator on the camera through the garage door. BUT, I did set up a night profile using the advice found herein, and I tested by opening the garage door temporarily to allow the camera's inbuilt illuminators to do their work. After a car passed by, I ran and checked the footage and found a near perfect license plate exposure on the first attempt!!
I obviously have some more work to do to get a more permanent camera placement for day/night use, but thanks to all the folks in these forums, and especially the superstar ALPR geniuses like MikeLud1, this looks super successful so far! I am continuing to do various tuning of zones and such to perfect the daytime collects, and will try to tackle Node-Red integration next.
THANK YOU!
I'll also add that CPAI really made this whole thing possible for me, and for many others, I suspect. I had tried in the past to deal with all the messy software installs and versioning hassles to get some AI guts working for this kind of task, whether on a Raspberry Pi or a Jetson Nano. But it proved to be a nightmare, and all the good tutorials proved to be almost useless the next time a software update came out for one of the MANY software packages involved, as the new updates invariably broke something that I was not qualifed to troubleshoot.
So, the Code Project AI folks are heroes in my book!!!
Attachments
Thanks @MikeLud1!
This approach works, but unfortunately not in the way I’d like. My goal is to ignore/cancel plates of certain parked cars within the field of view while still capturing plates of passing cars.
If the plates aren't in the "to cancel" list, BI reports the plates of both the parked and passing cars (when I only want to ignore the parked one).
However, if I add the plate of a parked car to the cancel list, BI cancels the entire alert—even if another (passing) plate is also detected in the alert.
It seems this might be a bug (or at least not a fully polished cancel feature) in BI.
If that’s the case, I'm surprised no one else has noticed or raised concerns. For instance, it also means if a bicycle is on the "to cancel" list and both a bicycle and car appear in an alert, BI won’t alert on the car either (even if car is set to be alerted).
I guess that's something to report to BI?
Guys sorry for reposting but not sure if my last post was missed. Having uninstalled and installed multiple times I am still getting the error message AI not responding at localhost 31268
CPAI is up and running (see below), custom models is populated (see below)
Just banging my head against the wall. Can anyone point me in the right direction?
CPAI is up and running (see below), custom models is populated (see below)
Just banging my head against the wall. Can anyone point me in the right direction?
No that didnt work. I also tried the actual IP address of the computer and that made no difference either.
Open a command prompt and enter the command netstat -na | find "32168"
You should see something like TCP 0.0.0.0:32168 0.0.0.0:0 LISTENING
Copy and paste the results into a reply
Also turn off your firewall or make sure you are allowing incoming TCP on port 32168
Here you go. Sorry I have no idea what this means
TCP 0.0.0.0:32168 0.0.0.0:0 LISTENING
This line means that some program is listening on port 32168 on all network interfaces.
The lines that say ESTABLISHED means that something established a connection to your machine.
Since the lines start with 127.0.0.1 (localhost) and the other end is 127.0.0.1 that means that you are making a successful connection to your own machine on the localhost interface.
The ones that show TIME_WAIT are old connections that will be closed. Probably from a previous browser session.
THe ones with [::1} are IPv6 and can most likely be ignored.
IN summary the program is running and listening on the correct port and something on your computer has made a successful connection to it and it all appears to be working correctly.
If your browser is not displaying the Web UI, I would next troubleshoot the browser itself, maybe a security setting or something.
Antivirus, javascript disabled, something like that.
You could also try installing another browser like chrome or firefox and trying that.
I have it all setup, just doing CPU recording (soon to test TPU), I see ALPR working, but I cannot play back the Alerts and see the ALPR info, it is in the video thumbnail and in the mouse over preview, but not in the main recorded video playback. Any idea what I might have disabled in this situation?
Can anyone explain why BI is confirming a lower-confidence license plate detection over one with a higher confidence level?
In the image below, there’s a detection with 95% confidence showing the correct plate (AIFD650), but BI has chosen a lower-confidence detection at 89% with an incorrect plate number (AIOFD650).
Shouldn't BI prioritize the highest-confidence detection when confirming the alert?
In the image below, there’s a detection with 95% confidence showing the correct plate (AIFD650), but BI has chosen a lower-confidence detection at 89% with an incorrect plate number (AIOFD650).
Shouldn't BI prioritize the highest-confidence detection when confirming the alert?
Vettester
Getting comfortable
- Feb 5, 2017
- 817
- 817
I see a couple of issues. First on the recording tab I would recommend setting your video option to “Continuous" instead of "Continuous sub + Alerts”. Also, you’ve got your AI misconfigured. The way you have your custom models set it will only identify what is in the "license-plate” which is DayPlatYoue or NightPlate so you won’t get actual tag numbers. I recommend you filter out everything except for the tags themselves and use just the LPR model. Here’s an example of one of my LPR cameras:
Also keep in mind that for this to work the “ALPR for plates” option (on the AI tab in your main BI settings) must be enabled.
Last edited:
Vettester
Getting comfortable
- Feb 5, 2017
- 817
- 817
