I haven't confirmed for the CPAI LPR, but when using Plate Recognizer, I use &PLATE.
See 5.6.8 - January 8, 2023
5.5.8 - June 13, 2022 - Code Project’s SenseAI Version 1 - See V2 here https://ipcamtalk.com/threads/codeproject-ai-version-2-0.68030/
- Thread starter fenderman
- Start date
I haven't confirmed for the CPAI LPR, but when using Plate Recognizer, I use &PLATE.
See 5.6.8 - January 8, 2023
Mike, where in Tapatalk is this new thread located? The original was under What's New.I started a new thread for all topic CodeProject.AI version 2.0.
CodeProject.AI Version 2.0
CodeProject.AI version 2.0 was released Jan 16, 2023, this thread is for all topic CodeProject.AI version 2.0. https://www.codeproject.com/Articles/5322557/CodeProject-AI-Server-AI-the-easy-way If you need any info on CodeProject.AI version 1.x please refer to the below link (there is some...ipcamtalk.com
I don't see the new thread yet.
MikeLud1; Did anything ever come of this testing for module(s) for older Nvidia GPUs?
I've been using a GT710, a Kepler based gpu, and I actually couldn't get it to work with Deepstack. The error indicated that the Pytorch version didn't support a card that old. It seems you would have to use CUDA 8.0 or something like that.
I've been trying version 2.6.0 several times, and the Yolo 5.6.2 is extremely slow, like 3 seconds or more for detection times. (CPU)
One time, I was able to get the Intel GPU to work, it was a little slower than version 1.6.8, like 500 to 700 msec.
Now I have gone back to Deepstack on my Nvidia Jetson. (Ipcam-combined or Ipcam-dark, getting 150 to 300 msec)
Just wondering if there was anything that was found about these older GPUs.
MikeLud1
IPCT Contributor
For your GT710 you need to install CUDA 10.2 and use Object Detection (YOLOv5 3.1) module
CUDA 10.2 link
CUDA Toolkit 10.2 Download
Get CUDA Toolkit 10.2 for Windows, Linux, and Mac OSX.
developer.nvidia.com
OK, I have been very successfully running the CPAI version 1.6.7 for a few months now on a "CPU only" (i5-8500) system, but am a bit hesitant to upgrade to the newer 2.0.x versions, partly due to my general confusion over which modules I would want to run. My current CPAI only loads the "Object Detection (YOLO)" module, with no further refinement of Version/Revision number. Now I see that we have several choices, even for CPU only systems (or those with integrated Intel Graphics), but how do we know which one to choose? Despite reading every post in all the relevant threads, I see no clear instructions for direction on this point.For your GT710 you need to install CUDA 10.2 and use Object Detection (YOLOv5 3.1) module
CUDA 10.2 link
CUDA Toolkit 10.2 Download
Get CUDA Toolkit 10.2 for Windows, Linux, and Mac OSX.
View attachment 152067
Also, I see some users are seeing slower analysis times compared to the 1.6.x version of CPAI: has this been resolved yet? I am usually one to jump into Beta versions fairly quickly, but this upgrade gave me pause for several reasons. Advice welcome...
MikeLud1
IPCT Contributor
For you Integrated Intel GPU I would try Object Detection (YOLOv5 .NET) and compare with Object Detection (YOLOv5 6.2) and see which one performs better, it should be Detection (YOLOv5 .NET)
Thanks Mike; I'll give it a try.For your GT710 you need to install CUDA 10.2 and use Object Detection (YOLOv5 3.1) module
CUDA 10.2 link
CUDA Toolkit 10.2 Download
Get CUDA Toolkit 10.2 for Windows, Linux, and Mac OSX.
View attachment 152067
Install the 10.* toolkit as well?
MikeLud1
IPCT Contributor
No joy so far.
Getting the following in the CPAI log:
2023-01-27 12:12:59: Object Detection (YOLOv5 3.1): Detecting using ipcam-combined in Object Detection (YOLOv5 3.1)
2023-01-27 12:12:59: Object Detection (YOLOv5 3.1): Unable to load model at C:\Program Files\CodeProject\AI\modules\YOLOv5-3.1\custom-models\ipcam-combined.pt (CUDA error: no kernel image is available for execution on the device) in Object Detection (YOLOv5 3.1)
Also, nothing works in the CPAI Explorer page for Object Detection or Custom Object Detection.
I will check and see if I have some pieces of CUDA 11 on the machine...
I downloaded CodeProject and the CPAI Explorer page for Object Detection or Custom Object Detection works for a minute and then it keeps popping up with a server connection error. I haven't gone any further with trying to incorporate into BI as I assume if this isn't working, then it won't work BI?
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: Pausing on error for 60 secs.
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
14:21:39: detect_adapter.py: Server connection error. Is the server URL correct?
14:21:39: detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
MikeLud1
IPCT Contributor
MikeLud1
IPCT Contributor
I am not sure what is happening can you post all details on the issue you are having on CodeProject.AI (link is below), they are very responsive and will look to fix it.
CodeProject.AI Discussions - CodeProject
www.codeproject.com
Odd. This is what I see.
I might need to dig through the registry.
I did this before to get rid of all traces of CUDA, I was having a similar error problem trying to run Deepstack in Docker in WSL.
As I recall, I had to look and look and delete everything I found until the error went away.
I keep going back to Deepstack, unfortunatly by changing the CUDA checkbox on the AI tab, that broke BI for a few hours.
Rebuilt the database, everything is working now.
Staying with Deepstack on the Nano. I could roll back to CPAI 1.6.8. That one worked well for about 6 weeks.
I might also hold off until the versions do not end with "beta".
MikeLud1
IPCT Contributor
Try the below steps
- remove CUDA 10.2
- delete all the CUDA folders
- remove all the CUDA paths
- reboot
- install CUDA 10.2
I am not sure what is happening can you post all details on the issue you are having on CodeProject.AI (link is below), they are very responsive and will look to fix it.
CodeProject.AI Discussions - CodeProject
So it is appearing the CodeProject needs internet connection to work?
I tried CodeProject on my old laptop that I rarely give internet access to.
Gave it internet access to install CodeProject and tested it and it works. Was returning the results with the test images.
Then took away the internet and it quit working.
Thinking maybe it was an issue with my install, I uninstalled it and reloaded from scratch but gave the computer internet access for several hours.
Every time I tried the option of testing with the test images it worked.
But then as soon as I disconnected internet, even though it would show the API server is online, it would spit out the errors in the log and not identify:
detect_adapter.py: Server connection error. Is the server URL correct?
detect_adapter.py: Pausing on error for 60 secs.
detect_adapter.py: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?objectdetection_queue: [ClientConnectorError] : Unable to check the command queue objectdetection_queue. Is the server URL correct?
So much for it being a local solution?
OK, I guess I have OCD about this. I keep saying I'm going to give up on this and just let the Jetson run it.
I'm reading that nvidia-smi will show what version of CUDA the Nvidia driver will support.
The command nvcc --version will show the current CUDA version installed.
Reference:
Different CUDA versions shown by nvcc and NVIDIA-smi
I am very confused by the different CUDA versions shown by running which nvcc and nvidia-smi. I have both cuda9.2 and cuda10 installed on my ubuntu 16.04. Now I set the PATH to point to cuda9.2. So...
stackoverflow.com
"I think I've seen this exact question come up multiple times over the last couple days. But I can't seem to find a duplicate now. The answer is: nvidia-smi shows you the CUDA version that your driver supports."
Like this:
Microsoft Windows [Version 10.0.19045.2486]
(c) Microsoft Corporation. All rights reserved.
C:\Users\ftsadmin>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:32:27_Pacific_Daylight_Time_2019
Cuda compilation tools, release 10.2, V10.2.89
C:\Users\ftsadmin>nvidia-smi
Sat Jan 28 07:52:08 2023
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 473.81 Driver Version: 473.81 CUDA Version: 11.4 |
|-------------------------------+----------------------+----------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... WDDM | 00000000:01:00.0 N/A | N/A |
| 50% 0C P0 N/A / N/A | 572MiB / 2048MiB | N/A Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
No running processes because I have CPAI service disabled
I have also checked all the paths for CUDA in user and system, they all point to CUDA 10.2, so that should be ok.
I'm going to check the driver version for CUDA 10.2 and install that Nvidia driver and see what that does.
I might also try to install the latest version of Blue Iris; reluctantly, because I have been sticking with critical or highly stable updates for the last year, I get fewer problems that way.
Last edited:
OK, here is where I am.OK, I guess I have OCD about this. I keep saying I'm going to give up on this and just let the Jetson run it.
I'm reading that nvidia-smi will show what version of CUDA the Nvidia driver will support.
The command nvcc --version will show the current CUDA version installed.
Reference:
Different CUDA versions shown by nvcc and NVIDIA-smi
I am very confused by the different CUDA versions shown by running which nvcc and nvidia-smi. I have both cuda9.2 and cuda10 installed on my ubuntu 16.04. Now I set the PATH to point to cuda9.2. So...
"I think I've seen this exact question come up multiple times over the last couple days. But I can't seem to find a duplicate now. The answer is: nvidia-smi shows you the CUDA version that your driver supports."
Like this:
Microsoft Windows [Version 10.0.19045.2486]
(c) Microsoft Corporation. All rights reserved.
C:\Users\ftsadmin>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:32:27_Pacific_Daylight_Time_2019
Cuda compilation tools, release 10.2, V10.2.89
C:\Users\ftsadmin>nvidia-smi
Sat Jan 28 07:52:08 2023
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 473.81 Driver Version: 473.81 CUDA Version: 11.4 |
|-------------------------------+----------------------+----------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... WDDM | 00000000:01:00.0 N/A | N/A |
| 50% 0C P0 N/A / N/A | 572MiB / 2048MiB | N/A Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
No running processes because I have CPAI service disabled
I have also checked all the paths for CUDA in user and system, they all point to CUDA 10.2, so that should be ok.
View attachment 152280
I'm going to check the driver version for CUDA 10.2 and install that Nvidia driver and see what that does.
I might also try to install the latest version of Blue Iris; reluctantly, because I have been sticking with critical or highly stable updates for the last year, I get fewer problems that way.
Changing the driver for the GT 710 had no effect.
Updating to CPAI version 2.07 had no effect.
Anytime I google the cuda error, I keep coming across the fact that the pytorch in cuda 10.2 only supports compute of 3.7 or higher, would have to compile pytorch again for a lower compute. I was seeing this more than a year ago when I was trying to run Deepstack GPU on this machine.
So:
Changed the Yolov5 3.1 back to cpu. Started testing with test images.
My favorite, pexels-ngrh-mei-5975635, gives me a round trip of between 480 to 500 ms.
I tried the Yolov5 6.2, with the same images, and I was getting 2 seconds +, which is comparable with what I was seeing running with Blue Iris.
So I am going to use Yolov5 3.1 in CPU mode and change Blue Iris back to using CPAI.
Bottom line is I think the GT 710 is a lost cause,
If anyone has actually successfully used the gpu in a GT 710 with CPAI, I would like to know.
Using CPAI with Yolov5 3.1 in CPU mode.
Done for now.