BI CPAI Loryta/Empire/Dahua IPC-B52IR-Z12E S2 LPR Config 12/2024

[seanstephens]

I have the IPC-B52IR-Z12E S2 camera and purchased the BI5-Full and installed CPAI 2.9.5 (LPR 3.3.4 - see attached). HW is i7-7700K and NVidia P600. Plenty of SSD

Watched "The Hook Up" video three times and went thru the setup per that video. The UI for BI and the camera have both changed since that video was posted a year ago so there is some confusion. I have reviewed many articles in this forum and BI Forum without success. Going on 10 hours now over two days and I am looking for some direction to get this specific combo working for Day/Night LPR capture, not to become an expert. BI Status shows blank AI tab. Also there are about 4 or 5 random recordings, probably during settings changes following along the video tutorial. Interestingly, the BI Camera AI settings/Static Object analysis recognizes DayPlates and vehicles (see attached)

Added screenshots for clarity of current setup. I didn't insert all the CamUI Image pages but those match all from the tutorial video by The Hook Up. Also, the BI Web Server, Schedule, and Location tabs all match the video recommendations. Any guidance would be greatly appreciated.

 

[seanstephens]

I tried about 50 different combinations from this forum and others and it's working well. MQTT is next!

 

[Flintstone61]

You can enable ONVIF/ camera events and disable Motion from BI.
The IVS rules inside the camera will sense more precisely the IVS motion trigger rules.
I'm just experimenting with IVS the last couple of weeks and I getting less false triggers. thanks to @wittaj.
You might have too many variables going with BI Motion, Codepreject motion, and internal Camera motion.

 

[Flintstone61]

Sometimes the distance is short enough to use the Z4e version of this bullet cam.
I find it less picky than a Zoomed Z12 in a location like you can see in one of your uploads.

 

[mpl]

Which model size do you use in cpai?

Nvidia p600 or p6000?

 

[seanstephens]

Thanks for your feedback!

Nvidia P600. CodeAI is using CPU, not GPU. Even though I enable GPU, LPR module restarts and reverts to CPU. See error log attached. Does CodeAI 2.9.5 not work with CUDA 12?

The motion detection and clarity are working now. Alerts are recording and I can play them and see the plates clearly, day or night. It seems the schedules were misconfigured.

But i still have a problem with the CodeAI LPR module "interpreting" the plates. Just lists "Motion_A", see attached. It sees them and records the clip but doesn't interprit, or list them. I installed latest CUDA (12.6.85)and cuDNN 9.6 to see if that would help. Same errors. Something about CUDA V10 and downloading "cudnn-10.0-windows10-x64-v7.6.5.32.zip from NVIDIA's official website"

From CodeAI My current System Info
Attached
My CodeAI server logs
Attached

 

[seanstephens]

Well, after breaking other items. I took these steps:
Visited CodeProject website and found article Aug 2024 referencing CUDA11.8/cuDNN 8.6.4 install. Installed 11.8 from Nvidia. Link to Nvidia cuDNN downloaded a script that no longer worked. Apparently Nvidia prevents DL of that older version of cuDNN. So I created an Nvidia acct and downloaded it from their archives. Downloaded 8.6.4 it and put proper files into toolkit subdirectories (C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.8\bin \inclide \lib\x64). Restarted CodeAI server and everything looks good now in CodeAI status (see attached).

Now BI5 isn't receiving any motion captures, which was working. the goal here after getting with BI support was to get rid of errors in CodeAI, which we have done. If I find a solution, I will post.

 

[seanstephens]

Restarted PC. starting to capture again but back to last problem, ALPR not recording plate numbers for search. I will post a dat file with screenshot for help

Notice the dat file tags the DayPlate in one of the frames but still lists the recording as Motion_A. I would like to capture the license plate so I can search and feed to MQTT

Here is the text output:
T-132 msec [7097 msec]

[
{
"api":"objects",
"found":{
"message":"Found truck,
car",
"count":2,
"predictions":[
{
"confidence":0.66357421875,
"label":"truck",
"x_min":0,
"y_min":101,
"x_max":121,
"y_max":298}
,
{
"confidence":0.9501953125,
"label":"car",
"x_min":448,
"y_min":190,
"x_max":1621,
"y_max":798}
]
,
"success":true,
"processMs":550,
"inferenceMs":532,
"moduleId":"ObjectDetectionYOLOv5-6.2",
"moduleName":"Object Detection (YOLOv5 6.2)",
"code":200,
"command":"detect",
"requestId":"0449d042-d282-4fb8-85f0-5eca89e0d358",
"inferenceDevice":"GPU",
"analysisRoundTripMs":1797,
"processedBy":"localhost",
"timestampUTC":"Fri,
10 Jan 2025 16:15:25 GMT"}
}
,
{
"api":"license-plate",
"found":{
"message":"Found DayPlate",
"count":1,
"predictions":[
{
"confidence":0.91259765625,
"label":"DayPlate",
"x_min":1449,
"y_min":641,
"x_max":1543,
"y_max":706}
]
,
"success":true,
"processMs":97,
"inferenceMs":95,
"moduleId":"ObjectDetectionYOLOv5-6.2",
"moduleName":"Object Detection (YOLOv5 6.2)",
"code":200,
"command":"custom",
"requestId":"f7876cea-fb02-4afb-94a9-a7bfea9bba50",
"inferenceDevice":"GPU",
"analysisRoundTripMs":202,
"processedBy":"localhost",
"timestampUTC":"Fri,
10 Jan 2025 16:15:25 GMT"}
}
,
{
"api":"alpr",
"found":{
"success":true,
"processMs":3277,
"inferenceMs":983,
"predictions":[
]
,
"message":"No plates found",
"moduleId":"ALPR",
"moduleName":"License Plate Reader",
"code":200,
"command":"alpr",
"requestId":"f3f9b549-7aba-4f20-aa8b-9e8e15e56752",
"inferenceDevice":"GPU",
"analysisRoundTripMs":5056,
"processedBy":"localhost",
"timestampUTC":"Fri,
10 Jan 2025 16:15:30 GMT"}
}
]