New CodeProject.AI Object Detection (YOLO11 .NET) Module

[nullex]

Try installing Cuda 12.4.0_551.61?

As i understand it, cuda is backwards compatible, so using the latest version of cuda (13 in my case) should work with any previous version of cuda.
But, because I'm running CPAI in docker, I think CUDA is handled by the container and not what's running on the host.
The issue with CPAI in docker is that pytorch isnt compiled or built for rtx 5000 series GPUs yet.

 

[MikeLud1]

As i understand it, cuda is backwards compatible, so using the latest version of cuda (13 in my case) should work with any previous version of cuda.
But, because I'm running CPAI in docker, I think CUDA is handled by the container and not what's running on the host.
The issue with CPAI in docker is that pytorch isnt compiled or built for rtx 5000 series GPUs yet.

I am working on building a Docker container that should work with RTX 5000 series GPUs. I will try to get the container released this weekend

 

[nullex]

I am working on building a Docker container that should work with RTX 5000 series GPUs. I will try to get the container released this weekend

That would be awesome and greatly appreciated. I've been messing around exec'd in to the running container trying to replace python 3.8 venv with 3.10 for yolov8 module to run newer versions of pytorch that support 5000 series but have not been having much luck.

 

[MikeLud1]

That would be awesome and greatly appreciated. I've been messing around exec'd in to the running container trying to replace python 3.8 venv with 3.10 for yolov8 module to run newer versions of pytorch that support 5000 series but have not been having much luck.

I just pushed mikelud/ai-server:cuda12_8, try and see if this works for you using the Object Detection (YOLOv5 .NET) 1.14.0 module, it should be using GPU (CUDA). I have the container running and my RTX 5090 works. The Object Detection (YOLOv5 6.2) 1.10.0 module still does not work, I need to still work on this. I just need to know if I am going in the right path.

https://hub.docker.com/r/mikelud/ai-server/tags

 

[nullex]

I just pushed mikelud/ai-server:cuda12_8, try and see if this works for you using the Object Detection (YOLOv5 .NET) 1.14.0 module, it should be using GPU (CUDA). I have the container running and my RTX 5090 works. The Object Detection (YOLOv5 6.2) 1.10.0 module still does not work, I need to still work on this. I just need to know if I am going in the right path.

https://hub.docker.com/r/mikelud/ai-server/tags

You rock!! YOLOv5 .NET is working great with GPU now!!

Thank you, this is great!!! Definitely feel free to let me know if I can help more in any way, I'd be happy to!

 

[wittaj]

I have been running an older CPAI version and feel it is time to make the jump to the latest and use these new modules.

I was going to delete my current one and install a fresh copy.

Do we have a cheat sheet on how/where to go about downloading the latest version and getting it running - is it Github now?

 

[Skinny1]

I'm glad users like @Vettester post snapshots of their screens. I installed the downloaded appsettings.json file, stopped and restarted the CP.ai service, and I still don't see the YOLO11 items.

Mike has a detailed explanation on how to install the new models in the front of this post.

 

[prsmith777]

I have been running an older CPAI version and feel it is time to make the jump to the latest and use these new modules.

I was going to delete my current one and install a fresh copy.

Do we have a cheat sheet on how/where to go about downloading the latest version and getting it running - is it Github now?

I'm still hanging out here in the back row watching things unfold. Had no time anyway last few weeks to change anything, but I have time off coming up and I'm close to moving forward.

I would till need the delivery model to work with this new setup before I proceed. Is that coming soon?

 

[Vettester]

Do we have a cheat sheet on how/where to go about downloading the latest version and getting it running - is it Github now?

I'm not aware of a detailed cheat sheet, but I've done this several times and can list the steps I take. The following steps are for a Windows install on a PC without a dedicated GPU. These steps assume you know how to navigate your way through Windows.

Remove existing CPAI installation

1. Open Windows Service (services.msc) and stop the CodeProject.AI.Server service
2. Open Control Panel, select Programs > Programs and Features and uninstall CodeProject.AI Server
3. Open Windows File Explorer and delete the following two directories
   • C:\Program Files\CodeProject
   • C:\ProgramData\CodeProject (this is a hidden directory)

Install Latest Version (2.9.5) of CPAI

1. Download CPAI and install from this LINK
2. After the install is complete open the web UI and stop any modules that are running
3. Open Windows Service (services.msc) and stop the CodeProject.AI.Server service

Modify CPAI Settings

1. Download Mike's updated appsettings.json file from this LINK
2. Open Windows File Explorer and navigate to C:\Program Files\CodeProject\AI\Server
3. Rename appsettings.json to appsettings.json.bac
4. Unzip the file downloaded in step 1 to C:\Program Files\CodeProject\AI\Server

Restart CPAI

1. Open Windows Service (services.msc) and start the CodeProject.AI.Server service

Install New Modules

1. Open the CPAI UI and navigate to the install Modules tab
2. Install the new YOLO Models that you want
   • Everything should be working now, if not restart the CPAI service again

 

[wittaj]

I'm not aware of a detailed cheat sheet, but I've done this several times and can list the steps I take. The following steps are for a Windows install on a PC without a dedicated GPU. These steps assume you know how to navigate your way through Windows.

Remove existing CPAI installation

1. Open Windows Service (services.msc) and stop the CodeProject.AI.Server service
2. Open Control Panel, select Programs > Programs and Features and uninstall CodeProject.AI Server
3. Open Windows File Explorer and delete the following two directories
   • C:\Program Files\CodeProject
   • C:\ProgramData\CodeProject (this is a hidden directory)

Install Latest Version (2.9.5) of CPAI

1. Download CPAI and install from this LINK
2. After the install is complete open the web UI and stop any modules that are running
3. Open Windows Service (services.msc) and stop the CodeProject.AI.Server service

Modify CPAI Settings

1. Download Mike's updated appsettings.json file from this LINK
2. Open Windows File Explorer and navigate to C:\Program Files\CodeProject\AI\Server
3. Rename appsettings.json to appsettings.json.bac
4. Unzip the file downloaded in step 1 to C:\Program Files\CodeProject\AI\Server

Restart CPAI

1. Open Windows Service (services.msc) and start the CodeProject.AI.Server service

Install New Modules

1. Open the CPAI UI and navigate to the install Modules tab
2. Install the new YOLO Models that you want
   • Everything should be working now, if not restart the CPAI service again

This is awesome! Thanks!

 

[MikeLud1]

I just pushed a new version of the docker container. This version now has the Object Detection (YOLO11 .NET) 1.3.0 module. This version supports both YOLO11 and YOLOv5 models. All of my YOLOv5 models are included until I can train the YOLO11 versions on the models. The YOLOv5 models have a suffix of -v5, to use them you would for example ipcam-general is now ipcam-general-v5.

There are two tags for the container, they are the same container, you can use ether one.

 

[nullex]

I just pushed a new version of the docker container. This version now has the Object Detection (YOLO11 .NET) 1.3.0 module. This version supports both YOLO11 and YOLOv5 models. All of my YOLOv5 models are included until I can train the YOLO11 versions on the models. The YOLOv5 models have a suffix of -v5, to use them you would for example ipcam-general is now ipcam-general-v5.

There are two tags for the container, they are the same container, you can use ether one.

This is awesome, thank you! yolo11 working great with my 5060ti in docker on linux vm. didn't try yolov5 6.2 but really happy to have yolo11!
Thank you so much, again!

 

[MikeLud1]

didn't try yolov5 6.2 but really happy to have yolo11!

The yolov5 6.2 module does not work so most likely I will be removing it. It is not needed because the YOLO11 module work with both YOLOv5 and YOLO11 models.