New CodeProject.AI License Plate Recognition (YOLO11) Module

Happy to help and reduce the burden on Mike, looking forward to that ipcam-combined, this new license plate reader is amazing.

Edit: Not to derail things because this is working perfectly for me, but the number of inference report counts is wonky. I restarted CPAI so I could use Mike's new appsettings (working) but instead of starting at 0 it started in the 7000's which is where it was when I shut it down, and it didn't resume counting from there, it flips back and forth between low and mid 7000's until I triggered a plate reading and then it took 15 images and scanned them all and started flipping between three numbers, the actual count and the old counts.

I just did a fresh install of windows server 2025 and windows 11 pro. Just installed geforce drivers and runtime and SDK drivers also blueiris and codeproject ai. Still getting the same problem with the new 1.0.4 yolo11. Maybe I am missing something?

I just updated to 1.0.4 from 1.0.3 and yolo11 stopped working again.

MikeLud1 said:

I just released Version 1.0.4, this version uses the original v1.0.0 way of initializing the models using default GPU Device ID 0. One addition is now you can also manually select the GPU Device ID as before v1.0.0 was fixed to GPU Device ID 0.
If the GPU does not initialize due to incorrect GPU Device ID it should automatically fallback to CPU. Check Task Manager and lookup you GPU ID and then go to CodeProject.AI Dashboard and under GPU Device Selection select the correct GPU Device ID

GPU initialized correctly and is using the GPU
View attachment 230593
View attachment 230594

GPU did not initialized correctly and is using the CPU
View attachment 230595
View attachment 230596

Click to expand...

OK, I tried starting from scratch again. I completed these steps in this order (as best I can remember):
1. Uninstalled all CodeProject AI modules.
2. Stopped CP.AI service.
3. Uninstalled CodeProject AI v 2.9.5 from control panel Programs and Features.
4. Deleted CodeProject AI folders in Program Files and Program Data.
5. Removed two CodeProject AI keys I found in Windows Registry.
6. Rebooted machine.
7. Installed CodeProject AI v 2.9.5 selecting only License Plate Reader v 3.3.4 and Object Detection (YOLOv5 .NET) v 1.14.0
At this point License Plate Reader v 3.3.4 was capturing license plate numbers
8. Replaced appsettings.json with new version.
9. Restarted CP.AI service and hit ctrl-F5.
10. Installed License Plate Reader (YOLO11-ONNX) v 1.0.4.
11. Stopped License Plate Reader v 3.3.4
12. Tried starting License Plate Reader (YOLO11-ONNX) v 1.0.4. Will not run.
13. Started License Plate Reader v 3.3.4
License Plate Reader v 3.3.4 is capturing license plate numbers

MikeLud1 said:

Can you post your CP.AI log file, sometimes iGPUs take some time to initialize

Click to expand...

I rolled back to 3.3.4 and then updated to 1.0.4 and it started working again.



Edit: Here's the log entry right before it failed.

2025-10-20 17:32:12: ** Module 'License Plate Recognition (YOLO11)' 1.0.4 (ID: ALPRYOLO11)
2025-10-20 17:32:12: ** Valid: True
2025-10-20 17:32:12: ** Module Path: <root>\modules\ALPRYOLO11
2025-10-20 17:32:12: ** Module Location: Internal
2025-10-20 17:32:12: ** AutoStart: True
2025-10-20 17:32:12: ** Queue: alpr_queue
2025-10-20 17:32:12: ** Runtime: python3.10
2025-10-20 17:32:12: ** Runtime Location: Local
2025-10-20 17:32:12: ** FilePath: alpr_adapter.py
2025-10-20 17:32:12: ** Start pause: 0 sec
2025-10-20 17:32:12: ** Parallelism: 0
2025-10-20 17:32:12: ** LogVerbosity:
2025-10-20 17:32:12: ** Platforms: windows
2025-10-20 17:32:12: ** GPU Libraries: installed if available
2025-10-20 17:32:12: ** GPU: use if supported
2025-10-20 17:32:12: ** Accelerator:
2025-10-20 17:32:12: ** Half Precision: enable
2025-10-20 17:32:12: ** Environment Variables
2025-10-20 17:32:12: ** CHAR_BOX_DILATION_HEIGHT = 3
2025-10-20 17:32:12: ** CHAR_BOX_DILATION_WIDTH = 0
2025-10-20 17:32:12: ** CHAR_CLASSIFIER_CONFIDENCE = 0.40
2025-10-20 17:32:12: ** CHAR_DETECTOR_CONFIDENCE = 0.40
2025-10-20 17:32:12: ** CORNER_DILATION_PIXELS = 5
2025-10-20 17:32:12: ** DEBUG_IMAGES_DIR = <root>\modules\ALPRYOLO11\debug_images
2025-10-20 17:32:12: ** DEVICE_ID = 0
2025-10-20 17:32:12: ** ENABLE_SPEED_DETECTION = True
2025-10-20 17:32:12: ** ENABLE_STATE_DETECTION = False
2025-10-20 17:32:12: ** ENABLE_VEHICLE_DETECTION = False
2025-10-20 17:32:12: ** MODELS_DIR = <root>\modules\ALPRYOLO11\models
2025-10-20 17:32:12: ** ONNX_MODELS_DIR = <root>\modules\ALPRYOLO11\models
2025-10-20 17:32:12: ** PLATE_ASPECT_RATIO = 2.5
2025-10-20 17:32:12: ** PLATE_DETECTOR_CONFIDENCE = 0.45
2025-10-20 17:32:12: ** SAVE_DEBUG_IMAGES = False
2025-10-20 17:32:12: ** SPEED_CENTROID_THRESHOLD = 5.0
2025-10-20 17:32:12: ** SPEED_IOU_THRESHOLD = 0.15
2025-10-20 17:32:12: ** STATE_CLASSIFIER_CONFIDENCE = 0.45
2025-10-20 17:32:12: ** USE_CUDA = False
2025-10-20 17:32:12: ** USE_DIRECTML = True
2025-10-20 17:32:12: ** USE_ONNX = True
2025-10-20 17:32:12: ** VEHICLE_CLASSIFIER_CONFIDENCE = 0.45
2025-10-20 17:32:12: ** VEHICLE_DETECTOR_CONFIDENCE = 0.45
2025-10-20 17:32:12:
2025-10-20 17:32:12: Started License Plate Recognition (YOLO11) module
2025-10-20 17:32:12: Running module using: launcher

i am getting this when trying to install the new module "Error in Install ALPRYOLO11: Response timeout. Try increasing the timeout value"

This is the windows 11 pro fresh install. Let me know if you need anything else.
Thanks Mike

it says installed but now i get this error...

19:42:46:Started License Plate Recognition (YOLO11) module
19:42:47:alpr_adapter.py: Traceback (most recent call last):
19:42:47:alpr_adapter.py: File "C:\Program Files\CodeProject\AI\modules\ALPRYOLO11\alpr_adapter.py", line 7, in
19:42:47:alpr_adapter.py: import cv2
19:42:47:alpr_adapter.py: ModuleNotFoundError: No module named 'cv2'
19:42:47:Module ALPRYOLO11 has shutdown

I even installed cuda and cudnn still same results.

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 lot more frequent. I'm seeing it happen about 3% of the time.

Thanks

No errors with dxdiag. I am using cuda 12.4 no issues with lpr 3.3.4. Is there any software needed besides NVidia drivers, cuda, CPAI 2.9.5, runtime 9.0.10 and sdk 9.0.306?
I uploaded a device manager picture it shows 3 displays 3060, intel 770 and remote display adaptor since i remote in, not sure that affects anything.

MikeLud1 said:

Before I released 1.0.0 during testing I was see this, I thought I fix it, can you send me the full resolution of the image without the plate read.

Click to expand...

Do you mean without burning in the plate in the image? If so, I'll have to disable that and send you the next one I find.