Blue Iris and CodeProject.AI ALPR
- Thread starter MikeLud1
- Start date
I think I have ALPR working partially. I now get this in the log, where can I find the plate info that it detected? Also, using the main stream check box really slows down my detection times. Is it necessary for ALPR? Thanks
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [ipcam-general] vehicle:94% [262,482 2008,1515] 3100ms
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [license-plate] DayPlate:91% [1550,1021 1762,1192] 3100ms
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [ipcam-general] vehicle:94% [262,482 2008,1515] 3100ms
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [license-plate] DayPlate:91% [1550,1021 1762,1192] 3100ms
What is the resolution of you camera, if it is 4MP you can try setting the substream to 1080p then use the substream
I'll try that on the two cameras I want to use ALPR on and see how it works. Where is the plate data stored?
This is all I get in the log.
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [ipcam-general] vehicle:94% [262,482 2008,1515] 3100ms
0 1/31/2023 5:54:45.141 PM Mailbox_Cam AI: [license-plate] DayPlate:91% [1550,1021 1762,1192] 3100ms
Try running a test like below to check if CP.AI ALPR module is work so we can tell if it is a BI or CP.AI issue.
23:36:56:Client request 'alpr' in the queue (...dedf80)
23:36:56:Request 'alpr' dequeued for processing (...dedf80)
23:36:56:License Plate Reader: Retrieved ALPR_queue command
23:36:56:Client request 'custom' in the queue (...b53dd0)
23:36:56:Request 'custom' dequeued for processing (...b53dd0)
23:36:56:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
23:36:57:Object Detection (YOLOv5 6.2): Detecting using license-plate
23:36:57:Response received (...b53dd0)
23:36:57:Object Detection (YOLOv5 6.2): Queue and Processing Object Detection (YOLOv5 6.2) command 'custom' (...b53dd0) took 78ms
23:36:57:ALPR_adapter.py: [2023/01/31 23:36:57] ppocr WARNING: Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process
23:36:57:ALPR_adapter.py: [2023/01/31 23:36:57] ppocr DEBUG: dt_boxes num : 1, elapse : 0.025981664657592773
23:36:57:License Plate Reader: [Exception] : Traceback (most recent call last):
23:36:57:Response received (...dedf80)
23:36:57:License Plate Reader: Queue and Processing License Plate Reader command 'alpr' (...dedf80) took 148ms
23:36:59:Client request 'alpr' in the queue (...23accd)
23:36:59:Request 'alpr' dequeued for processing (...23accd)
23:36:59:License Plate Reader: Retrieved ALPR_queue command
23:36:59:Client request 'custom' in the queue (...1d4fc0)
23:36:59:Request 'custom' dequeued for processing (...1d4fc0)
23:36:59:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
23:36:59:Object Detection (YOLOv5 6.2): Detecting using license-plate
23:36:59:Object Detection (YOLOv5 6.2): Queue and Processing Object Detection (YOLOv5 6.2) command 'custom' (...1d4fc0) took 81ms
23:36:59:Response received (...1d4fc0)
23:36:59:ALPR_adapter.py: [2023/01/31 23:36:59] ppocr WARNING: Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process
23:36:59:ALPR_adapter.py: [2023/01/31 23:36:59] ppocr DEBUG: dt_boxes num : 1, elapse : 0.03297924995422363
23:36:59:License Plate Reader: [Exception] : Traceback (most recent call last):
23:36:59:Response received (...23accd)
23:36:59:License Plate Reader: Queue and Processing License Plate Reader command 'alpr' (...23accd) took 158ms
Last edited:
Try restarting the service and run the test before any other AI calls, do the test twice. What GPU do you have and how much memory does it have.
You might be running out of GPU memory. Try replacing the license-plate model with the attached. After replacing the model restart CP.AI service. This model will use less GPU memory. After restarting the service rerun the ALPR test.
Thanks for sending, no luck unfortunately...
23:56:15:Request 'alpr' dequeued for processing (...67d88f)
23:56:15:Client request 'alpr' in the queue (...67d88f)
23:56:16:License Plate Reader: Retrieved ALPR_queue command
23:56:16:Client request 'custom' in the queue (...f8a342)
23:56:16:Request 'custom' dequeued for processing (...f8a342)
23:56:16:Object Detection (YOLOv5 6.2): Retrieved objectdetection_queue command
23:56:16:Object Detection (YOLOv5 6.2): Detecting using license-plate
23:56:16:Object Detection (YOLOv5 6.2): Queue and Processing Object Detection (YOLOv5 6.2) command 'custom' (...f8a342) took 53ms
23:56:16:Response received (...f8a342)
23:56:16:ALPR_adapter.py: [2023/01/31 23:56:16] ppocr WARNING: Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process
23:56:16:ALPR_adapter.py: [2023/01/31 23:56:16] ppocr DEBUG: dt_boxes num : 1, elapse : 0.024985074996948242
23:56:16:ALPR_adapter.py: [2023/01/31 23:56:16] ppocr WARNING: Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process
23:56:16:ALPR_adapter.py: [2023/01/31 23:56:16] ppocr DEBUG: dt_boxes num : 1, elapse : 0.02498459815979004
23:56:16:License Plate Reader: [Exception] : Traceback (most recent call last):
23:56:16:Response received (...67d88f)
23:56:16:License Plate Reader: [Exception] : Traceback (most recent call last):
23:56:16:License Plate Reader: Queue and Processing License Plate Reader command 'alpr' (...67d88f) took 172ms
I am not sure what is happening, below is a normal log if the ALPR is working. The bold line is not happening for you, it is failing. It still might be a GPU memory issue.
Can you open a command prompt and run nvidia-smi, then screenshot the results. This will show me how much GPU memory is being used.
24:04:46:Client request 'alpr' in the queue (...2ed341)
24:04:46:Request 'alpr' dequeued for processing (...2ed341)
24:04:46:License Plate Reader: Retrieved ALPR_queue command
24:04:46:Client request 'custom' in the queue (...fb5c73)
24:04:46:Request 'custom' dequeued for processing (...fb5c73)
24:04:46:Object Detection (Net): Command completed in 18 ms.
24:04:46:Response received (...fb5c73)
24:04:46:ALPR_adapter.py: [2023/02/01 00:04:46] ppocr WARNING: Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process
24:04:46:ALPR_adapter.py: [2023/02/01 00:04:46] ppocr DEBUG: dt_boxes num : 3, elapse : 0.028025150299072266
24:04:46:ALPR_adapter.py: [2023/02/01 00:04:46] ppocr DEBUG: rec_res num : 3, elapse : 0.012011051177978516
24:04:46:License Plate Reader: Queue and Processing License Plate Reader command 'alpr' (...2ed341) took 81ms
24:04:46:Response received (...2ed341)
It looks like it is a memory issue. I can make a lite version of the ALPR module that I hope can run on GPUs like yours. I will work with Chris from CP.AI to get one made or have the option to change to run as a lite version.
