|
What Operating system? Seems like a permissions issue
cheers
Chris Maunder
|
|
|
|
|
I followed a best practice guide for blue iris (the hookup). One of the recommendation was to use the custom model ipcam-combined for each camera. This setting was preventing CodeProject.AI automated installer from adding the Object Detection YOLOv5.NET and Object Detection YOLOv5 6.2 modules. After removing the custom model entries, the installer almost immediately began installing appropriately.
Details:
I was running 2.0.8 with that custom model ipcam-combined fine but every time I tried to upgrade it would not work. Multiple different versions, uninstalls, reinstalls, clean installs etc. BI AI would not automatically install the appropriate modules.
Hope this helps someone.
|
|
|
|
|
I upgraded from 2.4.8 to 2.5.1. After the upgrade I can no longer start any/all of the modules. I also noticed that Python is no longer running.
I've tried restarting the computer and re-installing CodeProject 2.5.1, neither one helped.
Quote: Error in Install ObjectDetectionYOLOv5-6.2: Call failed.
Quote: Error in Install FaceProcessing: Call failed.
System Info:
Quote:
☀️
Server version: 2.5.1
System: Windows
Operating System: Windows (Microsoft Windows 10.0.19045)
CPUs: Intel(R) Core(TM) i7-10700K CPU @ 3.80GHz (Intel)
1 CPU x 8 cores. 16 logical processors (x64)
GPU (Primary): NVIDIA GeForce GTX 1050 Ti (4 GiB) (NVIDIA)
Driver: 536.23, CUDA: 12.2 (up to: 12.2), Compute: 6.1, cuDNN:
System RAM: 32 GiB
Platform: Windows
BuildConfig: Release
Execution Env: Native
Runtime Env: Production
.NET framework: .NET 7.0.10
Default Python:
Video adapter info:
NVIDIA GeForce GTX 1050 Ti:
Driver Version 31.0.15.3623
Video Processor NVIDIA GeForce GTX 1050 Ti
System GPU info:
GPU 3D Usage 6%
GPU RAM Usage 1.1 GiB
Global Environment variables:
CPAI_APPROOTPATH = <root>
CPAI_PORT = 32168
Server logs:
Quote:
11:38:26:System: Windows
11:38:26:Operating System: Windows (Microsoft Windows 10.0.19045)
11:38:26:CPUs: Intel(R) Core(TM) i7-10700K CPU @ 3.80GHz (Intel)
11:38:26: 1 CPU x 8 cores. 16 logical processors (x64)
11:38:26:GPU (Primary): NVIDIA GeForce GTX 1050 Ti (4 GiB) (NVIDIA)
11:38:26: Driver: 536.23, CUDA: 12.2 (up to: 12.2), Compute: 6.1, cuDNN:
11:38:26:System RAM: 32 GiB
11:38:26:Platform: Windows
11:38:26:BuildConfig: Release
11:38:26:Execution Env: Native
11:38:26:Runtime Env: Production
11:38:26:.NET framework: .NET 7.0.10
11:38:26:Default Python:
11:38:26:App DataDir: C:\ProgramData\CodeProject\AI
11:38:26:Video adapter info:
11:38:26: NVIDIA GeForce GTX 1050 Ti:
11:38:26: Driver Version 31.0.15.3623
11:38:26: Video Processor NVIDIA GeForce GTX 1050 Ti
11:38:26:STARTING CODEPROJECT.AI SERVER
11:38:26:RUNTIMES_PATH = C:\Program Files\CodeProject\AI\runtimes
11:38:26:PREINSTALLED_MODULES_PATH = C:\Program Files\CodeProject\AI\preinstalled-modules
11:38:26:MODULES_PATH = C:\Program Files\CodeProject\AI\modules
11:38:26:PYTHON_PATH = \bin\windows\%PYTHON_NAME%\venv\Scripts\python
11:38:26:Data Dir = C:\ProgramData\CodeProject\AI
11:38:26:Server version: 2.5.1
11:38:26:ModuleRunner Start
11:38:27:Starting Background AI Modules
11:38:32:Server: This is the latest version
11:38:32:Current Version is 2.5.1
11:39:04:Call failed
11:56:37:Call failed
12:32:43:Call failed
12:32:49:Call failed
12:32:56:Call failed
12:32:59:Call failed
12:33:08:Call failed
12:33:10:Call failed
12:33:11:Call failed
12:33:13:Call failed
12:33:14:Call failed
12:33:16:Call failed
12:33:17:Call failed
13:20:22:Call failed
Logging level
|
|
|
|
|
Can you please force a non-cache update of the dashboard (eg. Ctrl+F5 in Windows on Chrome) and try again, and if there's an error, please post the error text.
Thanks,
Sean Ewington
CodeProject
|
|
|
|
|
I'm using the latest Firefox browser. When I performed the Ctrl-F5 and then tried to start the module it still failed but with a different error.
14:14:06:Preparing to install module 'ObjectDetectionYOLOv5-6.2'
14:14:06:Downloading module 'ObjectDetectionYOLOv5-6.2'
14:14:06:Unable to download module 'ObjectDetectionYOLOv5-6.2' from https://www.codeproject.com/KB/articles/5348853/ObjectDetectionYOLOv5-6.2-1.9.0.zip. Error: The SSL connection could not be established, see inner exception.
14:14:06:Unable to download module 'ObjectDetectionYOLOv5-6.2' from https://www.codeproject.com/KB/articles/5348853/ObjectDetectionYOLOv5-6.2-1.9.0.zip
I noticed a new tab called "Mesh", so I opened the tab and clicked on "Start". Then tried starting the module again. This time it WORKED.
Thanks for the help !!!
Current Server mesh status
GAME
Hostname: GAME
System: Windows (x64) NVIDIA GeForce GTX 1050 Ti
Platform: Windows
Active: true
Forwarding Requests: true
Accepting Requests: true
Visible Servers:
Routes Available: (4 processed)
vision/custom 0ms, 0 processed
vision/custom/list 0ms, 0 processed
vision/detection 150.5ms, 4 processed
Remote Servers in mesh: 0
Server logs
14:17:39:Starting mesh broadcasting
14:17:39:Starting mesh broadcast monitoring
14:17:39:Starting known mesh server pinging
14:18:22:Preparing to install module 'ObjectDetectionYOLOv5-6.2'
14:18:22:Downloading module 'ObjectDetectionYOLOv5-6.2'
14:18:22:Installing module 'ObjectDetectionYOLOv5-6.2'
14:18:23:ObjectDetectionYOLOv5-6.2: Installing CodeProject.AI Analysis Module
14:18:23:ObjectDetectionYOLOv5-6.2: ======================================================================
14:18:23:ObjectDetectionYOLOv5-6.2: CodeProject.AI Installer
14:18:23:ObjectDetectionYOLOv5-6.2: ======================================================================
14:18:23:ObjectDetectionYOLOv5-6.2: 798.4Gb of 952Gb available on
14:18:23:ObjectDetectionYOLOv5-6.2: General CodeProject.AI setup
14:18:23:ObjectDetectionYOLOv5-6.2: Creating Directories...Done
14:18:23:ObjectDetectionYOLOv5-6.2: GPU support
14:18:24:ObjectDetectionYOLOv5-6.2: CUDA Present...Yes (CUDA 12.2, No cuDNN found)
14:18:24:ObjectDetectionYOLOv5-6.2: ROCm Present...No
14:18:27:ObjectDetectionYOLOv5-6.2: Reading ObjectDetectionYOLOv5-6.2 settings.......Done
14:18:27:ObjectDetectionYOLOv5-6.2: Installing module Object Detection (YOLOv5 6.2) 1.9.0
14:18:27:ObjectDetectionYOLOv5-6.2: Installing Python 3.7
14:18:27:ObjectDetectionYOLOv5-6.2: Python 3.7 is already installed
14:18:27:ObjectDetectionYOLOv5-6.2: Creating Virtual Environment (Shared)...Virtual Environment already present
14:18:27:ObjectDetectionYOLOv5-6.2: Confirming we have Python 3.7 in our virtual environment...present
14:18:46:ObjectDetectionYOLOv5-6.2: Downloading Standard YOLO models...Expanding...Done.
14:18:46:ObjectDetectionYOLOv5-6.2: Copying contents of models-yolo5-pt.zip to assets...done
14:18:52:Response timeout. Try increasing the timeout value
14:19:08:ObjectDetectionYOLOv5-6.2: Downloading Custom YOLO models...Expanding...Done.
14:19:08:ObjectDetectionYOLOv5-6.2: Copying contents of custom-models-yolo5-pt.zip to custom-models...done
14:19:08:ObjectDetectionYOLOv5-6.2: Installing Python packages for Object Detection (YOLOv5 6.2)
14:19:08:ObjectDetectionYOLOv5-6.2: [0;Installing GPU-enabled libraries: If available
14:19:10:ObjectDetectionYOLOv5-6.2: Ensuring Python package manager (pip) is installed...Done
14:19:12:ObjectDetectionYOLOv5-6.2: Ensuring Python package manager (pip) is up to date...Done
14:19:12:ObjectDetectionYOLOv5-6.2: Python packages specified by requirements.windows.cuda.txt
14:19:14:ObjectDetectionYOLOv5-6.2: - Installing Pandas, a data analysis / data manipulation tool...Already installed
14:19:15:ObjectDetectionYOLOv5-6.2: - Installing CoreMLTools, for working with .mlmodel format models...Already installed
14:19:16:ObjectDetectionYOLOv5-6.2: - Installing OpenCV, the Open source Computer Vision library...Already installed
14:19:18:ObjectDetectionYOLOv5-6.2: - Installing Pillow, a Python Image Library...Already installed
14:19:19:ObjectDetectionYOLOv5-6.2: - Installing SciPy, a library for mathematics, science, and engineering...Already installed
14:19:20:ObjectDetectionYOLOv5-6.2: - Installing PyYAML, a library for reading configuration files...Already installed
14:19:22:ObjectDetectionYOLOv5-6.2: - Installing PyTorch, an open source machine learning framework...Already installed
14:19:23:ObjectDetectionYOLOv5-6.2: - Installing TorchVision, for working with computer vision models...Already installed
14:19:24:ObjectDetectionYOLOv5-6.2: - Installing Ultralytics YoloV5 package for object detection in images...Already installed
14:19:26:ObjectDetectionYOLOv5-6.2: - Installing Seaborn, a data visualization library based on matplotlib...Already installed
14:19:26:ObjectDetectionYOLOv5-6.2: Installing Python packages for the CodeProject.AI Server SDK
14:19:27:ObjectDetectionYOLOv5-6.2: Ensuring Python package manager (pip) is installed...Done
14:19:29:ObjectDetectionYOLOv5-6.2: Ensuring Python package manager (pip) is up to date...Done
14:19:29:ObjectDetectionYOLOv5-6.2: Python packages specified by requirements.txt
14:19:31:ObjectDetectionYOLOv5-6.2: - Installing Pillow, a Python Image Library...Already installed
14:19:32:ObjectDetectionYOLOv5-6.2: - Installing Charset normalizer...Already installed
14:19:33:ObjectDetectionYOLOv5-6.2: - Installing aiohttp, the Async IO HTTP library...Already installed
14:19:35:ObjectDetectionYOLOv5-6.2: - Installing aiofiles, the Async IO Files library...Already installed
14:19:36:ObjectDetectionYOLOv5-6.2: - Installing py-cpuinfo to allow us to query CPU info...Already installed
14:19:37:ObjectDetectionYOLOv5-6.2: - Installing Requests, the HTTP library...Already installed
14:19:43:ObjectDetectionYOLOv5-6.2: YOLOv5.1m summary: 391 layers, 21805053 parameters, 0 gradients
14:19:43:ObjectDetectionYOLOv5-6.2: Adding AutoShape...
14:19:43:ObjectDetectionYOLOv5-6.2: Self test: Self-test passed
14:19:43:ObjectDetectionYOLOv5-6.2: Module setup time 00:01:19.50
14:19:43:ObjectDetectionYOLOv5-6.2: Setup complete
14:19:43:ObjectDetectionYOLOv5-6.2: Total setup time 00:01:20.59
14:19:43:Module ObjectDetectionYOLOv5-6.2 installed successfully.
14:19:43:
14:19:43:Module 'Object Detection (YOLOv5 6.2)' 1.9.0 (ID: ObjectDetectionYOLOv5-6.2)
14:19:43:Valid: True
14:19:43:Module Path: <root>\modules\ObjectDetectionYOLOv5-6.2
14:19:43:AutoStart: True
14:19:43:Queue: objectdetection_queue
14:19:43:Runtime: python3.7
14:19:43:Runtime Loc: Shared
14:19:43:FilePath: detect_adapter.py
14:19:43:Pre installed: False
14:19:43:Start pause: 1 sec
14:19:43:LogVerbosity:
14:19:43:Platforms: all,!raspberrypi,!jetson
14:19:43:GPU Libraries: installed if available
14:19:43:GPU Enabled: enabled
14:19:43:Parallelism: 0
14:19:43:Accelerator:
14:19:43:Half Precis.: enable
14:19:43:Environment Variables
14:19:43:APPDIR = <root>\modules\ObjectDetectionYOLOv5-6.2
14:19:43:CUSTOM_MODELS_DIR = <root>\modules\ObjectDetectionYOLOv5-6.2\custom-models
14:19:43:MODELS_DIR = <root>\modules\ObjectDetectionYOLOv5-6.2\assets
14:19:43:MODEL_SIZE = Medium
14:19:43:USE_CUDA = True
14:19:43:YOLOv5_AUTOINSTALL = false
14:19:43:YOLOv5_VERBOSE = false
14:19:43:
14:19:43:Started Object Detection (YOLOv5 6.2) module
14:19:43:Installer exited with code 0
14:19:44:Module ObjectDetectionYOLOv5-6.2 started successfully.
14:19:47:detect_adapter.py: Running init for Object Detection (YOLOv5 6.2)
14:23:15:Object Detection (YOLOv5 6.2): Rec'd request for Object Detection (YOLOv5 6.2) command 'detect' (...11e974) took 697ms
14:25:15:Object Detection (YOLOv5 6.2): Rec'd request for Object Detection (YOLOv5 6.2) command 'detect' (...0d55e0) took 247ms
14:27:14:Object Detection (YOLOv5 6.2): Rec'd request for Object Detection (YOLOv5 6.2) command 'detect' (...2f6864) took 258ms
14:31:13:Object Detection (YOLOv5 6.2): Rec'd request for Object Detection (YOLOv5 6.2) command 'detect' (...f7629d) took 273ms
14:34:30:Object Detection (YOLOv5 6.2): Rec'd request for Object Detection (YOLOv5 6.2) command 'detect' (...5619e0) took 253ms
|
|
|
|
|
that is truly bizarre. Mesh has nothing to do with the installer. However: whatever works! I'm glad you're back up and running.
(My guess: network caching was making your machine think you had an unreachable target, but starting mesh kicked off a cache refresh or something and it took another look and was good)
cheers
Chris Maunder
|
|
|
|
|
I install the new version (2.5.1).
I see this message, if I push the test button in Agent - Settings - AI Servers - Configure menu:
Result: OK
But in Object Detection (Coral) menu Test Result is this: AI test failed: ObjectDetectionCoral test not provisioned
But I see this in the Codeproject.AI Dashboard:
19:27:24:Object Detection (Coral): Retrieved objectdetection_queue command 'detect'
19:27:25:Object Detection (Coral): Rec'd request for Object Detection (Coral) command 'detect' (...c959a2) took 139ms
Agent v5.2.4.0
Before, I use the older Codeproject.AI 2.3.2, it's worked.
Please help me! Thanks!
|
|
|
|
|
Thanks very much for the report. When you test Object Detection (Coral) in the CodeProject.AI Server Explorer, does it work? Do you get any errors?
Thanks,
Sean Ewington
CodeProject
|
|
|
|
|
In the CodeProject.AI Server Explorer its work flawlessly. (143ms)
|
|
|
|
|
AgentDVR log.txt:
23:29:54 TmrBroadcastElapsed: Removed 1 rtc sessions
23:31:54 CAM1: SetFailed: Unable to perform detection at CoreLogic.AI.ObjectRecognizer.<detect>d__6.MoveNext()
23:31:54 CAM1: SetFailed: Will retry Object Recognition in 30 seconds.
23:31:54 CAM1: OpenWriter: StartSaving
23:31:54 CAM1: .ctor: Using stream timestamps for this recording
23:31:54 Open: OPEN RECORD
23:31:54 Open: written header
23:31:54 CAM1: Open: Recording (Raw Writer CoreLogic.Sources.Combined.RawMonitor)
23:31:54 CAM1: WarnClock: Dropped packet as out of order - set Use System Clock on recording tab if you have problems
23:31:55 CAM1: RecorderRecordingOpened: Recording Opened
23:32:33 CAM1: RecorderRecordingClosed: Recording Closed
23:32:33 CAM1: Close: Record stop
23:32:40 CAM1: SetFailed: Unable to perform detection at CoreLogic.AI.ObjectRecognizer.<detect>d__6.MoveNext()
23:32:40 CAM1: SetFailed: Will retry Object Recognition in 30 seconds.
23:32:41 CAM1: OpenWriter: StartSaving
23:32:41 CAM1: .ctor: Using stream timestamps for this recording
23:32:41 Open: OPEN RECORD
23:32:41 Open: written header
23:32:41 CAM1: Open: Recording (Raw Writer CoreLogic.Sources.Combined.RawMonitor)
23:32:41 CAM1: WarnClock: Dropped packet as out of order - set Use System Clock on recording tab if you have problems
23:32:41 CAM1: RecorderRecordingOpened: Recording Opened
23:32:46 CAM2: SetFailed: Unable to perform detection at CoreLogic.AI.ObjectRecognizer.<detect>d__6.MoveNext()
23:32:46 CAM2: SetFailed: Will retry Object Recognition in 30 seconds.
23:32:46 CAM2: OpenWriter: StartSaving
23:32:46 CAM2: .ctor: Using stream timestamps for this recording
23:32:46 Open: OPEN RECORD
23:32:46 Open: written header
23:32:46 CAM2: Open: Recording (Raw Writer CoreLogic.Sources.Combined.RawMonitor)
23:32:46 CAM2: RecorderRecordingOpened: Recording Opened
|
|
|
|
|
"Unable to perform detection" is an error message coming back from CPAI.
Check the CPAI logs after Agent tries to send an image in.
|
|
|
|
|
I see this in the Codeproject.AI Dashboard, if I switch the logging level to trace:
07:11:43:Object Detection (Coral): Retrieved objectdetection_queue command 'detect'
07:11:43:Response received (#reqid f65e9036-0dc6-4d7d-a8ac-042d2b2c02a2 for command detect): No objects found
07:11:43:Object Detection (Coral): Rec'd request for Object Detection (Coral) command 'detect' (...2c02a2) took 135ms
07:11:44:Client request 'detect' in queue 'objectdetection_queue' (...ba08ac)
07:11:44:Request 'detect' dequeued from 'objectdetection_queue' (...ba08ac)
07:11:44:Object Detection (Coral): Retrieved objectdetection_queue command 'detect'
07:11:44:Response received (#reqid 56d0fd97-59ab-4e36-baa5-2eead4ba08ac for command detect): No objects found
07:11:44:Object Detection (Coral): Rec'd request for Object Detection (Coral) command 'detect' (...ba08ac) took 138ms
Even if its looks good, the AgentDVR said: AI down. So the problem still persist.
|
|
|
|
|
in Agent DVR if you go to Server Settings - Logging and set logging to Debug it'll show you the response from CPAI in the logs.
|
|
|
|
|
Let me know if you get this sorted out or if there's something our end you need fixed.
cheers
Chris Maunder
|
|
|
|
|
07:02:19 Cam1: Process: Recognize Objects
07:02:19 Cam1: Detect: {"message":"No objects found","count":0,"predictions":[],"success":true,"processMs":133,"inferenceMs":128,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":200,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24025,"failedInferences":79,"numInferences":24104,"numItemsFound":514,"averageInferenceMs":128.10131113423517,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":139,"processedBy":"localhost"}
07:02:20 Cam2: Process: Recognize Objects
07:02:20 Cam1: Process: Recognize Objects
07:02:20 Cam1: Detect: {"success":false,"predictions":[],"message":"","error":"Unable to perform detection","count":0,"processMs":4,"inferenceMs":0,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":500,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24025,"failedInferences":80,"numInferences":24105,"numItemsFound":514,"averageInferenceMs":128.10131113423517,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":9,"processedBy":"localhost"}
07:02:20 Cam1: Failed: AI Failure count at 1
07:02:20 Cam1: SetFailed: Unable to perform detection at CoreLogic.AI.ObjectRecognizer.<Detect>d__6.MoveNext()
07:02:20 Cam1: SetFailed: Will retry Object Recognition in 30 seconds.
07:02:20 Cam2: Detect: {"message":"No objects found","count":0,"predictions":[],"success":true,"processMs":134,"inferenceMs":129,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":200,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24026,"failedInferences":80,"numInferences":24106,"numItemsFound":514,"averageInferenceMs":128.10134853908266,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":140,"processedBy":"localhost"}
07:02:20 StartSaving: From Alert: True, From AI Alert: False
07:02:20 Cam1: OpenWriter: StartSaving
07:02:20 Cam1: .ctor: Using stream timestamps for this recording
07:02:20 Open: OPEN RECORD
07:02:20 Open: written header
07:02:20 Cam1: Open: Recording (Raw Writer CoreLogic.Sources.Combined.RawMonitor)
07:02:20 Cam1: RecorderRecordingOpened: Recording Opened
07:02:21 Cam2: Process: Recognize Objects
07:02:21 Cam2: Detect: {"message":"No objects found","count":0,"predictions":[],"success":true,"processMs":131,"inferenceMs":126,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":200,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24027,"failedInferences":80,"numInferences":24107,"numItemsFound":514,"averageInferenceMs":128.10126108128355,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":136,"processedBy":"localhost"}
07:02:22 Cam2: Process: Recognize Objects
07:02:22 Cam2: Detect: {"message":"No objects found","count":0,"predictions":[],"success":true,"processMs":132,"inferenceMs":127,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":200,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24028,"failedInferences":80,"numInferences":24108,"numItemsFound":514,"averageInferenceMs":128.10121524887631,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":137,"processedBy":"localhost"}
Really interesting.
Sometimes it works, sometimes it doesn't.
|
|
|
|
|
definitely an error on CPAI's side of things
Detect: {"success":false,"predictions":[],"message":"","error":"Unable to perform detection","count":0,"processMs":4,"inferenceMs":0,"moduleId":"ObjectDetectionCoral","moduleName":"Object Detection (Coral)","code":500,"command":"detect","executionProvider":"TPU","canUseGPU":false,"statusData":{"successfulInferences":24025,"failedInferences":80,"numInferences":24105,"numItemsFound":514,"averageInferenceMs":128.10131113423517,"histogram":{"baseball bat":10,"bird":3,"person":115,"microwave":24,"umbrella":1,"boat":5,"train":10,"car":288,"bench":28,"sheep":2,"knife":3,"toilet":1,"airplane":2,"cat":1,"suitcase":6,"fire hydrant":1,"bus":1,"clock":1,"cell phone":7,"dog":2,"mouse":2,"traffic light":1}},"analysisRoundTripMs":9,"processedBy":"localhost"}
|
|
|
|
|
I'm hoping to have a new version of the Coral module out today that will provide better error information for this case.
cheers
Chris Maunder
|
|
|
|
|
So what can I do? The problem is still presist. Thank you guys!
|
|
|
|
|
I assume that 2.5.1 is an actual production version since it is not marked as -alpha, -beta, rc etc.
Congratulations!
Installed the new version, and the install went smoothly.
Installed the Coral module, and that module is working well with Blue Iris and my M.2 Coral accelerator.
Thanks for all the hard work.
P.S. The Coral response times are faster by a factor of 2? i.e. where I was getting ~ 120 - 150 ms with ver 2.4.7, I am now getting 70 to 90 ms with 2.5.1.
|
|
|
|
|
Great, good to hear! Do you have the multi-TPU code enabled?
|
|
|
|
|
I only have the one TPU. I don't know what enabling that would do.
No.
Module 'Object Detection (Coral)' 2.1.1 (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
LogVerbosity:
Platforms: all
GPU Libraries: installed if available
GPU Enabled: enabled
Parallelism: 1
Accelerator:
Half Precis.: enable
Environment Variables
CPAI_CORAL_MULTI_TPU = false
CPAI_CORAL_USE_YOLO = true
MODELS_DIR = <root>\modules\ObjectDetectionCoral\assets
MODEL_SIZE = medium
Status Data: {
"successfulInferences": 578,
"failedInferences": 225,
"numInferences": 803,
"numItemsFound": 673,
"averageInferenceMs": 76.33391003460207,
"histogram": {
"car": 545,
"person": 76,
"truck": 44,
"bus": 7,
"train": 1
}
}
Started: 26 Jan 2024 8:24:55 AM Central Standard Time
LastSeen: 26 Jan 2024 9:24:52 AM Central Standard Time
Status: Started
Requests: 1372 (includes status calls)
Provider: TPU
CanUseGPU: False
HardwareType: GPU
|
|
|
|
|
Interesting, the non-multi-TPU code should be the same as before. I wonder if Chris will have any suggestions. If you enable the multi-TPU code you’re likely to see better concurrency and throughput, but I wouldn’t expect your timing numbers to improve. I wonder if the model that you are running is the same?
modified 26-Jan-24 11:04am.
|
|
|
|
|
I was going to blame you, Seth, for the speed kick but I actually suspect it's due to the use of the edgeTPU library rather than simply using the pycoral python code. Another possibility is he's using a different model than previous. There's a huge speed difference between MobileNet and EfficientDet.
cheers
Chris Maunder
|
|
|
|
|
Yeah. That’s the funny thing: I don’t think the new code will report numbers all that much better than the old code. It’s still doing the same things in the same order.
The new code will, however, handle a bajillion concurrent inferences. While one thread is running inference, another can be resizing. And another can be interpreting the results. And yet another can be running on a second TPU. And if someone is running with tiling, they might even be seeing slower speeds as it’s doing twice (or more) as much inference work in parallel.
Pillow-simd, on the other hand, will give better numbers. Maybe you’re right about the edgeTPU library? Maybe it’s compiled better than the default? Or maybe it’s just a different model.
|
|
|
|
|
Phase of the moon. I swear...
cheers
Chris Maunder
|
|
|
|
|