Blue Iris and CodeProject.AI ALPR

silvertonesx24 said:
Do you mean the Camera's zone map? I made that full screen and the yellow box still tracks "behind" the car and the plate. It's like a search area that isn't big enough
Click to expand...
Have you tried creating two zones (left third and right two-thirds) and using a crossing from right to left as the trigger to search for a plate?

I agree that more zoom would help give the license plate model a better chance to see a plate.
 
VideoDad said:
Have you tried creating two zones (left third and right two-thirds) and using a crossing from right to left as the trigger to search for a plate?

I agree that more zoom would help give the license plate model a better chance to see a plate.
Click to expand...

I'll give that a shot, the issue seems to be that the yellow overlay box that shows you motion tracking, doesn't cover the front license plates for oncoming cars. I'll zoom in a bit on the street, my concern with too much zoom though is that the car will pass too quickly and then miss the track
 
silvertonesx24 said:
I'll give that a shot, the issue seems to be that the yellow overlay box that shows you motion tracking, doesn't cover the front license plates for oncoming cars. I'll zoom in a bit on the street, my concern with too much zoom though is that the car will pass too quickly and then miss the track
Click to expand...
When AI is looking at the image it is not just looking at the section that has motion (the yellow box). It looks at the whole image/zone minus the sections that are optionally configured to be blacked out (which is maybe what you did with the non-road sections there on the top left and bottom right).

I agree with Vettester in that you need to zoom in more.

Open the video for playback and turn on "Analyze with AI" and you will see what I mean. Don't forget that even if it finds a plate it needs to be big enough to read it. I can't even read a single digit on the plate in you example. You look at Vettester's image and you can read the whole plate.
 
Last edited:
AlwaysSomething said:
When AI is looking at the image it is not just looking at the section that has motion (the yellow box). It looks at the whole image/zone minus the sections that are optionally configured to be blacked out (which is maybe what you did with the non-road sections there on the top left and bottom right).

I agree with Vettester in that you need to zoom in more.

Open the video for playback and turn on "Analyze with AI" and you will see what I mean. Don't forget that even if it finds a plate it needs to be big enough to read it. I can't even read a single digit on the plate in you example. You look at Vettester's image and you can read the whole plate.
Click to expand...
OK, I do have plenty of zoom on this, but it won't capture both sides of the street then
I'll have to buy another
 
prsmith777 said:
The link for the cuDNN install script seems to be dead:

File Download - CodeProject

www.codeproject.com www.codeproject.com

Is there a new link somewhere? I need to install for CUDA 11.8
Click to expand...
You can save the bat file for the below link
github.com

CodeProject.AI-Server/src/SDK/Scripts/install_CUDnn.bat at v2.6.5 · codeproject/CodeProject.AI-Server

CodeProject.AI Server is a self contained service that software developers can include in, and distribute with, their applications in order to augment their apps with the power of AI. - codeproject...
github.com github.com
 
  • Like
Reactions: prsmith777
MikeLud1 said:
You can save the bat file for the below link
github.com

CodeProject.AI-Server/src/SDK/Scripts/install_CUDnn.bat at v2.6.5 · codeproject/CodeProject.AI-Server

CodeProject.AI Server is a self contained service that software developers can include in, and distribute with, their applications in order to augment their apps with the power of AI. - codeproject...
github.com github.com
Click to expand...
Thanks. I noticed it says this script is for version 2.6.1. Im running 2.9.3. Will it still work correctly?
 
I'm working on getting 2.9.5 with the current ALPR working, and it seems I have mostly issues with trigger timing. I am foggy on the application of the setting "Analyze One Image Each: XXmsec". Does this control the time between each image, centered 'before' and 'after' the trigger point? ...or is it something else like how long to allow CPAI to analyze before cancelling an alert?

TIA,
Kev
 
truglo said:
I'm working on getting 2.9.5 with the current ALPR working, and it seems I have mostly issues with trigger timing. I am foggy on the application of the setting "Analyze One Image Each: XXmsec". Does this control the time between each image, centered 'before' and 'after' the trigger point? ...or is it something else like how long to allow CPAI to analyze before cancelling an alert?

TIA,
Kev
Click to expand...
It's right there in the Help PDF... it's the image sampling interval.

1757285654726.png
 
Last edited:
  • Like
Reactions: AP514, truglo and looney2ns
Hi everyone. It looks like this info has been here a while but I wanted to check if this is still needed as I am getting this error in CPAI

15:47:26:ALPR_adapter.py: W0923 15:47:26.203171 28580 dynamic_loader.cc:285] Note: [Recommend] copy cudnn into CUDA installation directory.
15:47:26:ALPR_adapter.py: For instance, download cudnn-10.0-windows10-x64-v7.6.5.32.zip from NVIDIA's official website,

Are people loading cudnn to get ALPR working? Ty
 
I would like to use CPAI. I suspect I am just not using the correct stuff or have it set up correctly. There is also a lot of older info out there so it becomes confusing given the cutting edge nature of this stuff. I appreciate any help you can provide. I have a NVIDIA card that can use CUDA so I would also like to be able to use CUDA for CPAI not just for license but also for AI of people and other stuff. I can see that I don't have cuDNN installed

Server version: 2.9.5
System: Windows
Operating System: Windows (Windows 11 24H2)
CPUs: Intel(R) Core(TM) i7-14700K (Intel)
1 CPU x 20 cores. 28 logical processors (x64)
GPU (Primary): NVIDIA GeForce RTX 4060 Ti (8 GiB) (NVIDIA)
Driver: 581.29, CUDA: 12.2.91 (up to: 13.0), Compute: 8.9, cuDNN:
System RAM: 128 GiB
Platform: Windows
BuildConfig: Release
Execution Env: Native
Runtime Env: Production
Runtimes installed:
.NET runtime: 9.0.9
.NET SDK: 9.0.305
Default Python: Not found
Go: Not found
NodeJS: Not found
Rust: Not found
Video adapter info:
Microsoft Remote Display Adapter:
Driver Version 10.0.26100.5074
Video Processor
NVIDIA GeForce RTX 4060 Ti:
Driver Version 32.0.15.8129
Video Processor NVIDIA GeForce RTX 4060 Ti
System GPU info:
GPU 3D Usage 0%
GPU RAM Usage 1.3 GiB
Global Environment variables:
CPAI_APPROOTPATH = <root>
CPAI_PORT = 32168

I was going to go to here
developer.nvidia.com

cuDNN 9.13.0 Downloads

developer.nvidia.com developer.nvidia.com
And get the latest. I am also not sure then if the custom model ipcam-combind is going to play well with things like Yolov8 which looks to be for modern GPU's (Which I think mine fits).

Again, thank you for any help
 
Just a follow up. I was able to add the cuDNN to the path and now it shows up. Still would like feed back on Yolov people are using now and wonder if v8 has custom models to use. I tried using the oned from v31. and v6.2 and they didn't work....
 
sdennis said:
Just a follow up. I was able to add the cuDNN to the path and now it shows up. Still would like feed back on Yolov people are using now and wonder if v8 has custom models to use. I tried using the oned from v31. and v6.2 and they didn't work....
Click to expand...
Use the Object Detection (YOLOv5 .NET) module it will use your Nvidia GPU and it is faster then all the other YOLO modules
 
  • Like
Reactions: Skinny1
You must log in or register to reply here.
Top Bottom