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/

I did not uninstall DS. below are the step I did for the change

  1. In BI main AI settings I disabled Auto start/stop, stopped DS Service, and disabled the AI
  2. Install CUDA 11.7.1
  3. Ran cuDNN script
  4. Installed CodeProject.AI 1.5.6.2
  5. After the install I went back into BI main AI settings and enabled AI, changed the port to 5000, and enabled Auto start/stop
  6. Since BI is not fully integrated I had to change the custom model folder in the registry to "C:\Program Files\CodeProject\AI\AnalysisLayer\CustomDetection\assets"
  7. The last step is to change the model names in each of the cameras. You can find all the custom model names in "C:\Program Files\CodeProject\AI\AnalysisLayer\CustomDetection\assets"
1661050968387.png
 
Nope!

"No predictions returned"

I've tried uninstalling and re-installing from scratch and it still does not work.

Edit: If I do 'detect scene' it comes up "Downtown: 52%"
Any suggestions? I’ve tried to uninstall, reboot. Tired repairing and rebooting etc. all the same outcome as per above.
 
Sorry but I've spent too much time on this already. Blue iris is in a state of transition from using DeepStack to Code Porject AI.
I've uninstalled DeepStack, I re-instlled DS, I've uninstalled DS, I've checked my global AI configurations and AI on individual cameras and my configurations are consistent with what others have posted.

I've 6 green boxed in CodeProject, I'm using the latest software, I used the bat script to install CUDnn, though I find it a bit of a challenge using W10 to run a batch file as someone more used to Ubuntu.

One odd thing I found a few days ago that in Program Files I had lost the NVIDIA directory, it was there before. NVIDIA Corporation and NVIDIA GPU Computing Toolkit were there. Strange because in the past when I was installing DeepStack I put the CUDnn stuff in the NVIDIA directory. Anyway I saw there was an updated Nvidia driver for my T600 GPU which once installed seems to have brought back the NVIDIA directory with >CUDNN>v8.5>bin,include,lib,zlib dated 20th August.

I have no working AI in BI but I don't really want to be putting DeepStack back on.
I'm going to lay down in a quiet dark room for now. Actually I'm going out for a walk.
 
It appears that most people here are having issues with the GPU install. Unless you really have a large install or need for it I would just go with the CPU method and spend your money on a fast CPU. These times now are really low for CPU and in most cases would work just fine for most and save yourself some headaches & electricity to boot :) Just for kicks I'm going to throw AI 1.5.6.2 on my demo and see what that 12th gen cpu can do, be right back with results :)
 
Revisiting the batch install script, install_CUDnn.bat, my W10 PC throws up an error when attempting to add in Environmental Variables
:: 3. Add this path to the PATH environment variable:
:: setx /M PATH = Path + "%PATH%;C:\Program Files\NVIDIA\CUDNN\v8.5\bin"

It doesn't like the forwardslash, the same I expect would happen with 4, because there's another forwardslash, a second /M.
 
  • Wow
Reactions: sebastiantombs
Well this test surprized me. The lower CPU did better with timings, go figure. The demo is using strictly motion detection where my main is constant record. I also had other stuff running on my I7 12700 machine at the time. But my point is both work very well. It took me a few more minutes because I had to bring my demo up to version as well. 5.6.0.3

Screenshot 2022-08-21 100923.png


Screenshot 2022-08-21 100816.png
 
Revisiting the batch install script, install_CUDnn.bat, my W10 PC throws up an error when attempting to add in Environmental Variables
:: 3. Add this path to the PATH environment variable:
:: setx /M PATH = Path + "%PATH%;C:\Program Files\NVIDIA\CUDNN\v8.5\bin"

It doesn't like the forwardslash, the same I expect would happen with 4, because there's another forwardslash, a second /M.
The two lines do not get executed they are commented
1661096923768.png
 
I am not locating a clear document (despite googling) that explains how to configure bi and cpai(codeproject ai).

Fresh windows 11 install of bi. No deepstack added. Latest release of cpai says its running.
Local host 5000 configured.

I see messages in bi saying motion detected and coming back nothing found. Nothing is being tagged for my clip review.

Aside from providing screen shots (not at pc currently) is there a step by step tutorial that i am over looking?

I can see cpai has many green boxes in the console. It says its running.

Thanks for any assistance in advanced.
 
I am not locating a clear document (despite googling) that explains how to configure bi and cpai(codeproject ai).

I've not see a guide, but a wealth of info in the thread. I got it working pretty good on my i7-6700 (no Nvidia card), Win 10 setup. Also, posting this for feedback. If I'm doing something wrong please let me know, but so far things are working much better than Deepstack every did (probably did not have that setup optimally). :)

Everything disabled except Custom Objects (using Medium resolution, the default). I tried High resolution, but slowed things down and not sure there was a benefit from it on my system). Edited the modulesettings.json to disable modules (found here C:\Program Files\CodeProject\AI\AnalysisLayer\ See pic, each folder has it own settings.json file.

Edit:forgot a pic.

Screenshot 2022-08-21 163740.jpg

Use Notepad++ Change activate to false.
Screenshot 2022-08-21 194421.jpg


Screenshot 2022-08-21 161054.jpg

My current BI settings. I had to edit the registry to point to the custom model folder and turn it on.
Screenshot 2022-08-21 161305.jpg
Screenshot 2022-08-21 165707.jpg

Screenshot 2022-08-21 161414.jpg
Screenshot 2022-08-21 161435.jpg

Edit: Removed vehicle and changed to car since its not part of ipcam-combined.
Screenshot 2022-08-21 161506.jpg
Screenshot 2022-08-21 161927.jpg
 
Last edited:
If I'm doing something wrong please let me know,
I see you are using the ipcam-combined custom model. I don't think it recognizes vehicles specifically as a label, you'll need to use car, truck, bus etc. to confirm. If you want to use vehicle as a label you'll need to use the ipcam-general model.

 
I see you are using the ipcam-combined custom model. I don't think it recognizes vehicles specifically as a label, you'll need to use car, truck, bus etc. to confirm. If you want to use vehicle as a label you'll need to use the ipcam-general model.


Thanks for the feedback. Your right vehicle is not recognized under that model. I had some cut and paste errors in there too. :) Using car does work. BTW is the dog based off of large dogs? I have Chihuahua and they are not labeled (movement is picked up though).
 
^ Have you read through this thread, plenty of useful info that should get you going.
All 17 pages of the thread, no sir.
I have been watching it off an on for about a week. There are bits n pieces in this and that post but assembling them into a easily digestable format for a relatively new AI user is where I am at.
I had DP AI working fine thanks to various users posting.
Since my win10 hdd died I figured I might as well give CPAI a shot.
Not as intuitive as has been hoped.
 
Thanks for the feedback. Your right vehicle is not recognized under that model. I had some cut and paste errors in there too. :) Using car does work. BTW is the dog based off of large dogs? I have Chihuahua and they are not labeled (movement is picked up though).
Thank you for posting your settings sir.
It helped confirm most of what I had already.

I am still at a loss how to actually edit the json to turn off the various modules running by CPAI. Right now I have 6 green boxes showing as started.
I too only need (think anyway) "Custom Object Detection" & "Object Detection (Python)" running.
Do not use the other items.
At some point I will take another stab at figuring out how to only have the 2 things enabled that I believe I use.

The other (important) item that helped was finally finding MikeLud's post about editing the registry to enable and change the path for the custom models.
Prior to that, I saw folks say "I hacked the registery as noted previously in the forums". lol, not super useful if you were not watching and noting such info prior. :)


Specifically his post here
#343
He forgot to mention he also edited the registry just above the path because that is the "enable" button that is greyed out in BI.
Change the value from 0 to 1 to enable the use of the custom folder.

Cant wait to see where this is going.