MikeLud1
IPCT Contributor
add alpr:0 to your camera's Custom Models, this will stop the alpr from running.
Blue Iris and CodeProject.AI ALPR
- Thread starter MikeLud1
- Start date
woolfman72
Getting the hang of it
- Joined
- Dec 28, 2015
- Messages
- 152
- Reaction score
- 73
look into installing a test loop back adapter its built into windows and see if that's fixes the issue. "localhost" and no internet has been a bane of my existence recently at work with some software.
Are you running NVIDIA CUDA or a built-in Intel GPU? Those response times seem really long. I'm running an NVIDIA GeForce GT 1030 with 2 GiB of RAM and mostly get sub 100ms responses, with the slowest ones still around 500ms. I think I'm on the latest version of BI and CPAI and I'm only using license-plate on LPR cameras and only using Ipcam-general on cameras where I want to detect vehicles but not plates.
Nvidia CUDA. I have a quadro K2200 4 GB Ram. I can’t seem to get the response times everybody else is getting. Using ALPR and Ipcam-general as well
Last edited:
Well what do you know - we learn something new everyday!
This fixed it!!!!
For those having same problem, these are the steps to install the microsoft loopback adapter:
- right click on window start menu icon and select Device manager. Device manager window will immediately open (or you may use any other way how to open device manager window)
- click on Action, and select Add legacy hardware
- click Next on welcome screen
- choose "Install the hardware that i manually select from a list" and click on Next
- scroll down and select Network adapters from offered common hardware types and click on Next
- select Microsoft as the manufacturer, and then select Microsoft KM-TEST Loopback adapter card model, click on Next
- click on Next
- click on Finish
MikeLud1
IPCT Contributor
I will let them know of the issue.
Vettester
Getting comfortable
- Joined
- Feb 5, 2017
- Messages
- 743
- Reaction score
- 698
I’m not seeing this in my log files. I wonder if now that you have your loop back issue resolved if reinstalling or repairing your CPAI would help?
When you pulled internet you are not getting the "error checking for available modules: No such host is known. (www.codeproject.com443)"
I just tried repair and that didn't do it, so I uninstalled and reinstalled and that didn't do it either.
Vettester
Getting comfortable
- Joined
- Feb 5, 2017
- Messages
- 743
- Reaction score
- 698
Just once.
Even with "Use custom models" checked, it is still grayed out. I tried checking every option but it doesn't enable it.
MikeLud1
IPCT Contributor
Try the good old reboot the computer, also make sure you are on the current version of BI
Vettester
Getting comfortable
- Joined
- Feb 5, 2017
- Messages
- 743
- Reaction score
- 698
The BI app cannot find your custom model folder. Since you're running CPAI on a remote server try clicking on the three dots and browse to the custom model folder on the remote machine. You can also keep the auto start/stop option disabled. Ken told me BI cannot start and stop services on a remote machine.
MikeLud1
IPCT Contributor
A workaround to get custom models to work is create a folder named custom models and add empty txt files with the custom model name and pt extension, so the file name should look like this
ipcam-general.pt. Then point BI to this new folder
Further to my previous message about rules for British plates, I have been giving the issue some thought.
If we are going with a generic license plate model then we can't bake the logic into the AI rules, as plates from elsewhere in Europe may have different rules. With that in mind I am working on some general UK rules that I am running as a post process to BI reporting the plate, so the plate is "corrected" before being passed to my downstream systems.
The biggest culprits seem to be "I" and "O".
I is easy to deal with as the letter I does not exist, so a global search and replace to convert to 1 works well.
"O" is the tricky one. As far as I can see there is no case where the number 0 would be in position 1 or 2 on the plate, so I do a conversion on the first 2 positions to replace any 0 (zero) with the letter O.
Once these processes have been run I check for a "modern" plate layout, so "AA??AAA" and replace any "O" with 0 (zero) in the number portion of the plate.
A lot of this does presume that you have a clear read of the plate and mistakes can still be made, but I am currently running on approx 97% accuracy rate during the day and alightly less at night, but still very good.
So unless we looked to have a UK localised AI solution that could be trained on the rules I think a lot of this work has to be done post-process rather than in the AI.
Anyway, that's my brain-dump for a monday morning, would appreciate thoughts.
If we are going with a generic license plate model then we can't bake the logic into the AI rules, as plates from elsewhere in Europe may have different rules. With that in mind I am working on some general UK rules that I am running as a post process to BI reporting the plate, so the plate is "corrected" before being passed to my downstream systems.
The biggest culprits seem to be "I" and "O".
I is easy to deal with as the letter I does not exist, so a global search and replace to convert to 1 works well.
"O" is the tricky one. As far as I can see there is no case where the number 0 would be in position 1 or 2 on the plate, so I do a conversion on the first 2 positions to replace any 0 (zero) with the letter O.
Once these processes have been run I check for a "modern" plate layout, so "AA??AAA" and replace any "O" with 0 (zero) in the number portion of the plate.
A lot of this does presume that you have a clear read of the plate and mistakes can still be made, but I am currently running on approx 97% accuracy rate during the day and alightly less at night, but still very good.
So unless we looked to have a UK localised AI solution that could be trained on the rules I think a lot of this work has to be done post-process rather than in the AI.
Anyway, that's my brain-dump for a monday morning, would appreciate thoughts.
^I would certainly agree with that logic. As I joked previously, unless @MikeLud1 clones himself, chatGPT can take over, or it turns into a paid solution, I think there are so many different local naming rules that it would be a monumental task to incorporate all of this in the AI.
I mean we know it can be done as Rekor and Plate Recognizer do it, but those are paid services.
I think most would be happy with this being able to simply read the characters with high enough accuracy that it could be used if something happened. So if 1 is mistaken for an I and a 0 for a O, it is still enough for the police to go on with the variations to try.
That is awesome you can write code to process this for you for your locale!
I mean we know it can be done as Rekor and Plate Recognizer do it, but those are paid services.
I think most would be happy with this being able to simply read the characters with high enough accuracy that it could be used if something happened. So if 1 is mistaken for an I and a 0 for a O, it is still enough for the police to go on with the variations to try.
That is awesome you can write code to process this for you for your locale!
Looking for a second set of fresh eyes on my settings. I must be missing something because I'm not getting any interaction between BI and CPAI as for as the license-plate model goes. Let me know if you see anything that's not correct. These settings are for my Mailbox cam and license-plate isn't showing up in the log. Thanks!
MikeLud1
IPCT Contributor
It could be you are running to many models, how are the times if you only run one model