|
I'm getting somehow the same issue that renders BlueIris to generate an AI Error 500.
Using Coral USB TPU on a Proxmox LXC CPAI Docker.
The only difference in my case is that the process restart itself eventually so I don't have to do it.
13:02:23:objectdetection_coral_adapter.py: WARNING:root:No work in 120.0 seconds, watchdog shutting down TPU runners!
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944063 252967 pipelined_model_runner.cc:172] Thread: 124925159392832 receives empty request
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944077 252967 pipelined_model_runner.cc:245] Thread: 124925159392832 is shutting down the pipeline...
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944214 252967 pipelined_model_runner.cc:255] Thread: 124925159392832 Pipeline is off.
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944253 266018 pipelined_model_runner.cc:207] Queue is empty and `StopWaiters()` is called.
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944337 252967 pipelined_model_runner.cc:172] Thread: 124925159392832 receives empty request
13:02:23:objectdetection_coral_adapter.py: E20240310 17:02:23.944345 252967 pipelined_model_runner.cc:240] Thread: 124925159392832 Pipeline was turned off before.
13:02:23:objectdetection_coral_adapter.py: I20240310 17:02:23.944375 252967 pipelined_model_runner.cc:207] Queue is empty and `StopWaiters()` is called.
13:02:23:objectdetection_coral_adapter.py: E20240310 17:02:23.944383 252967 pipelined_model_runner.cc:240] Thread: 124925159392832 Pipeline was turned off before.
13:02:23:objectdetection_coral_adapter.py: E20240310 17:02:23.944387 252967 pipelined_model_runner.cc:147] Failed to shutdown status: INTERNAL: Pipeline was turned off before.
<pre>Module 'Object Detection (Coral)' 2.1.4 (ID: ObjectDetectionCoral)
Valid: True
Module Path: <root>/modules/ObjectDetectionCoral
AutoStart: True
Queue: objectdetection_queue
Runtime: python3.9
Runtime Loc: Local
FilePath: objectdetection_coral_adapter.py
Pre installed: False
Start pause: 1 sec
Parallelism: 1
LogVerbosity:
Platforms: all
GPU Libraries: installed if available
GPU Enabled: enabled
Accelerator:
Half Precis.: enable
Environment Variables
CPAI_CORAL_MODEL_NAME = YOLOv8
CPAI_CORAL_MULTI_TPU = true
MODELS_DIR = <root>/modules/ObjectDetectionCoral/assets
MODEL_SIZE = small
Status Data: {
"inferenceDevice": "Multi-TPU",
"inferenceLibrary": "TF-Lite",
"canUseGPU": "false",
"successfulInferences": 53,
"failedInferences": 490,
"numInferences": 543,
"averageInferenceMs": 23.132075471698112
}
Started: 10 Mar 2024 2:44:04 PM Coordinated Universal Time
LastSeen: 10 Mar 2024 5:32:49 PM Coordinated Universal Time
Status: Started
modified 14-Mar-24 12:17pm.
|
|
|
|
|
I get the same error with multi tpu enabled. If I turn it off, it goes away.
modified 14-Mar-24 12:17pm.
|
|
|
|
|
Good,
I'll give it a try.
Thanks
modified 14-Mar-24 12:17pm.
|
|
|
|
|
After disabling Multi-Tpu support I'm now in the same boat. After a while AI stop responding and the only way is to restart the module.
modified 14-Mar-24 12:17pm.
|
|
|
|
|
Can you please try uninstalling and reinstalling the Coral module? We've applied a fix (we hope).
Thanks,
Sean Ewington
CodeProject
modified 14-Mar-24 12:17pm.
|
|
|
|
|
Have the exact same problem. Goes down multiple times a day. The Docker container should be catching this and restarting itself, but it's not.
|
|
|
|
|
So far the latest update fixed the issue for me
|
|
|
|
|
Hasn't for me, but I've heard they're working on an update to auto-restart when it goes down, so that's cool
|
|
|
|
|
Successfully installed LPR 3.0.2 with no errors however no plates are detected when queries are passed to CPAI from Agent.
Loading CPAI Explorer and processing images of license plates returns no predictions.
I have face detection also running on this VM with no issues.
Server version: 2.5.6
System: Windows
Operating System: Windows (Microsoft Windows 10.0.19045)
CPUs: AMD FX(tm)-6300 Six-Core Processor (AMD)
1 CPU x 2 cores. 2 logical processors (x64)
GPU (Primary): VirtualBox Graphics Adapter (WDDM) (Oracle Corporation)
Driver: 6.0.18.0
System RAM: 4 GiB
Platform: Windows
BuildConfig: Release
Execution Env: Native
Runtime Env: Production
Runtimes installed:
.NET runtime: 7.0.10
.NET SDK: Not found
Default Python: Not found
Go: Not found
NodeJS: Not found
Video adapter info:
VirtualBox Graphics Adapter (WDDM):
Driver Version 6.0.18.0
Video Processor VirtualBox VESA BIOS
System GPU info:
GPU 3D Usage 0%
GPU RAM Usage 0
Global Environment variables:
CPAI_APPROOTPATH = <root>
CPAI_PORT = 32168
19:54:11:Module ALPR installed successfully.
19:54:11:
19:54:11:Module 'License Plate Reader' 3.0.2 (ID: ALPR)
19:54:11:Valid: True
19:54:11:Module Path: <root>\modules\ALPR
19:54:11:AutoStart: True
19:54:11:Queue: alpr_queue
19:54:11:Runtime: python3.9
19:54:11:Runtime Loc: Local
19:54:11:FilePath: ALPR_adapter.py
19:54:11:Pre installed: False
19:54:11:Start pause: 3 sec
19:54:11:Parallelism: 0
19:54:11:LogVerbosity:
19:54:11:Platforms: all
19:54:11:GPU Libraries: installed if available
19:54:11:GPU Enabled: enabled
19:54:11:Accelerator:
19:54:11:Half Precis.: enable
19:54:11:Environment Variables
19:54:11:AUTO_PLATE_ROTATE = True
19:54:11:MIN_COMPUTE_CAPABILITY = 6
19:54:11:MIN_CUDNN_VERSION = 7
19:54:11:OCR_OPTIMAL_CHARACTER_HEIGHT = 60
19:54:11:OCR_OPTIMAL_CHARACTER_WIDTH = 30
19:54:11:OCR_OPTIMIZATION = True
19:54:11:PLATE_CONFIDENCE = 0.7
19:54:11:PLATE_RESCALE_FACTOR = 2
19:54:11:PLATE_ROTATE_DEG = 0
19:54:11:
19:54:11:Started License Plate Reader module
19:54:11:Installer exited with code 0
19:54:15:Module ALPR started successfully.
modified 7-Mar-24 10:23am.
|
|
|
|
|
Does it work if you run a test image like the below
|
|
|
|
|
No it doesn't detect anything
|
|
|
|
|
Are you seeing any errors in the log? Also what Object Detection module are you using.
|
|
|
|
|
When I upload an image and select read license plates the log returns a single line of:
16:33:59:Response rec'd from License Plate Reader command 'alpr' (...d722a0)
|
|
|
|
|
Are you running Object Detection (YOLOv5 6.2) or Object Detection (YOLOv5 .NET) module? Because the LPR module needs one of these modules to work.
|
|
|
|
|
Ah, now that would make sense! I did not have any object detection running, thought I was fine as I had already had face detection successfully running on the same host.
Have started the .NET module and now license plate detection is working. Apologies for bothering you.
|
|
|
|
|
|
I have the Yolo V5 6.2 running and it works perfect with BI.
But LPR doesn't work. I tested several times now. Always "No Predictions Returned". Yolo V5 6.2 returns "DayPlate found" and marks it correctly if I run it with the license-plate model.
During installation I found an error because of CUDNN. CUDA 12.2 is installed and Yolo V5 6.2 is running in GPU mode.
Is it mandatory for LPR to install CUDNN? I skipped it because "Face Processing" and "Yolo V5 6.2" didn't need it.
|
|
|
|
|
It is mandatory to have cuDNN installed also depending on the age of your GPU CUDA 12.2 will not work and needs CUDA 11.8
modified 7-Mar-24 9:08am.
|
|
|
|
|
Currently I'm using "platerecognizer". But BI doesn't deal with the results since a couple of months now. The only result is "DayPlate" but although the text is returned correctly, it fails to cancel the alerts for my own car. In a former version this worked too.
Now I want to test this CPAI solution which is much better, especially because I don't like to send such data into cloud storages.
I have cuDNN installer downloaded. Will test it later.
Here is the system info because of the GPU:
|
|
|
|
|
With your GPU I recommend using CUDA 11.8 not CUDA 12.0
|
|
|
|
|
I got it to run with cuDNN installation. But all sample plates were read with wrong results.
Unfortunately currently not reliable to detect at least German license plates with my 5 MP cameras.
"platerecognizer.com" can read it, but here BI fails with "cancel alert" rule as mentioned above. Can't verify if LPR would work if the results of test images aren't correct.
|
|
|
|
|
Try uninstalling the License Plate Reader module then reinstall.
|
|
|
|
|
The installation was ok, any errors in log about cuDNN or CUDA. It worked, but the results were wrong. "Badges" from admission office, "blanks" and "-" are detected as 8, ...
And the samples are really good from my 5 MP cameras. I also enabled OCR improvement, no changes.
It seems not to work with German license plates.
Send you PM with samples.
|
|
|
|
|
Got a new error message in latest version, one I don't recall seeing before. Not crashing level of problem but thought I would share...
08:19:37:Object Detection (YOLOv5 6.2): [RuntimeError] : Traceback (most recent call last):
File "C:\Program Files\CodeProject\AI\modules\ObjectDetectionYOLOv5-6.2\detect.py", line 141, in do_detection
det = detector(img, size=640)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\yolov5\models\common.py", line 705, in forward
y = self.model(x, augment=augment) # forward
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\yolov5\models\common.py", line 515, in forward
y = self.model(im, augment=augment, visualize=visualize) if augment or visualize else self.model(im)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\Lib\site-packages\yolov5\models\yolo.py", line 209, in forward
return self._forward_once(x, profile, visualize) # single-scale inference, train
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\Lib\site-packages\yolov5\models\yolo.py", line 121, in _forward_once
x = m(x) # run
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "C:\Program Files\CodeProject\AI\runtimes\bin\windows\python37\venv\Lib\site-packages\yolov5\models\yolo.py", line 75, in forward
wh = (wh * 2) ** 2 * self.anchor_grid[i] # wh
RuntimeError: The size of tensor a (60) must match the size of tensor b (56) at non-singleton dimension 2
modified 5-Mar-24 11:42am.
|
|
|
|
|
Thanks very much for your report. Could you please share your System Info tab from your CodeProject.AI Server dashboard?
Thanks,
Sean Ewington
CodeProject
|
|
|
|