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 don't want to derail this thread but Is anyone using SenseAi with an LPR camera at night? If so, which model are you using? I've tried both the license-plate and YOLOv5l models with mixed results. Hoping someone has a better suggestion.
 
I don't want to derail this thread but Is anyone using SenseAi with an LPR camera at night? If so, which model are you using? I've tried both the license-plate and YOLOv5l models with mixed results. Hoping someone has a better suggestion.
I am using my DeepStack ALPR. I am going to make CodeProject.AI ALPR module, hopefully will have it finished by the end of October maybe earlier.

1662698501782.png
1662698608202.png
 
I am going to make CodeProject.AI ALPR module, hopefully will have it finished by the end of October maybe earlier.
Thanks Mike, I was using your ALPR module with deepstacks before I switched to SenseAI. However, it was missing some plates at night which is one of the reasons I dumped deepstacks. I look forward to seeing how well your CodeProject.AI ALPR module works.
 
  • Like
Reactions: MikeLud1
Quick update - BI 5.6.0.8 and Codeproject 1.5.6-0002 GPU mode on a GTX 970 is working flawless. ~42K processed without a restart of BI or AI (since install 11am BST 04/09/2022), GPU VRAM holding steady at 0.9-1.1GB - Codeproject RAM steady at 1.8GB - I run 7 cameras all with substream for AI - average processing time 80ms, lows of 30ms highs of 160ms - most are around ~50-60ms - I'm happy with the switch to Codeproject on my 970 GTX system but on my 1650 GTX system I had to revert to Deepstack.

1662710354335.png1662709640494.png1662709715539.png
 

Attachments

  • 1662709533889.png
    1662709533889.png
    6.8 KB · Views: 9
Gentlemen, many thanks for providing me with the benefit of your knowledge.
MikeLud1, I’ve now found the path to the custom models - thank you
slidermike, excellent links! could you identify the line to stop the custom model folder path being greyed out please - the “deepstack_custom” ”value data” is already a “1” in my registry
actran, it’s been so long ago now since I dabbled with the .dat files I have forgotten how to use them in testing and tuning. See below. I will spend time on this.
View attachment 139325
Dave, right now you have to edit the windows registry to change the custom models folder location. It will remain greyed out in the BI gui until there is better integration.
Change the path to match where mike shows. Thats the default install location for CPAI (containers are diff) and with the bit set to "1" that tells windows to enable.
 
Dave, right now you have to edit the windows registry to change the custom models folder location. It will remain greyed out in the BI gui until there is better integration.
Change the path to match where mike shows. Thats the default install location for CPAI (containers are diff) and with the bit set to "1" that tells windows to enable.
Hi slidermike, I did change the custompath to precisely the same as MikeLud1 (the entire registry screen is identical) and checked the assets files were there in explorer but it still returns nothing found. I’ve now tried uninstalling CodeProject, reinstalling DeepStack and re-editing the registry but get the same nothing found result and the blue iris path error. But I can’t actually find similar object files in DeepStack (ipcam-general). I’ve spent hours and hours with no success but unfortunately don’t have the skill to identify and fix whatever I screwed up.

If I can figure out how to do it, I’m now thinking of taking the PC back to its out-of-the-box state and starting again with everything afresh. It’s a 10th generation PC with windows 11 that I use with RDP. There’s no risk I’ll lose my windows licence is there?

Perhaps waiting to install AI when CodeProject is integrated with Blue Iris would be wise in my case.
 
  • Like
Reactions: gwminor48
Gentlemen, I really appreciate your support! I had uninstalled all AI and deleted everything related I could find. But because you have taken the time to respond I have again installed CodeProject and re-edited the registry. Still nothing found but perhaps this time a lot closer to working... - I haven't had green ticks before.

Screenshot 2022-09-10 115805.pngScreenshot 2022-09-10 121243.pngScreenshot 2022-09-10 121500.pngScreenshot 2022-09-10 124101.pngScreenshot 2022-09-10 124311.png
Screenshot 2022-09-10 131408.png

Apart from the path failure popup, I notice a failure in the RegSaveKey in the log. It's not a valid path but I don't know if it matters or how to find it to edit.
Yet again, I'm hoping for your help.
 
@Dave Lonsdale

Dave,

Your "to confirm:" field is empty. You need to put the object(s) in there that you are looking to alert on such as vehicle and/or person

I might also suggest analyze more images. It looks like that van has not even made it fully into the frame by the time the analysis was over

1662813829064.png
 
Last edited:
@Dave Lonsdale

Dave,

I have had nothing but problems with BI starting CodeAI so I just unchecked that option and I am relying in the fact that it runs as a windows service

1662814375514.png
 
  • Like
Reactions: Tinman
The "Could not start; check path" information message appears to be a bug. The fact that you're seeing the Stop now vs the Start now option tells me that it is running.
Screen Shot 2022-09-10 at 7.06.17 AM.png
Like @clk8 mentioned, your problem is that you have left the "To confirm:" field blank.
 
Last edited:
Hello guys - I'm learning things all the time - like services.msc - very useful thanks.
So far as I know, when using the custom AI models and "default object detection" is unchecked, "to confirm" is left blank. I tried populating it with car etc and I get red failure crosses (see below).
I think I only got one image to process in the previous attached because there was a very high confidence in the alert analysed, even though it was a very high confidence of "nothing found". I discovered a long time ago with Deepstack that the way to analyse all the images is to put something in "To cancel" that will never be found (see below).
So, my problem has not yet been solved but great to get your suggestions. Thank you all. Any more please?
Is there a reason why CodeProject gets to the assets folders but apparently not their image contents? It did when I first installed CodeProject.

CodeProject is running:-
Screenshot 2022-09-10 193630.png

Analysing all the images:-
Screenshot 2022-09-10 191625.png

Objects in To Confirm when default is not checked:-
Screenshot 2022-09-10 191914.png

Clarity of the wanted alert image is not the problem:-
Screenshot 2022-09-10 190235.png
 
Hello guys - I'm learning things all the time - like services.msc - very useful thanks.
So far as I know, when using the custom AI models and "default object detection" is unchecked, "to confirm" is left blank. I tried populating it with car etc and I get red failure crosses (see below).
I think I only got one image to process in the previous attached because there was a very high confidence in the alert analysed, even though it was a very high confidence of "nothing found". I discovered a long time ago with Deepstack that the way to analyse all the images is to put something in "To cancel" that will never be found (see below).
So, my problem has not yet been solved but great to get your suggestions. Thank you all. Any more please?
Is there a reason why CodeProject gets to the assets folders but apparently not their image contents? It did when I first installed CodeProject.

CodeProject is running:-
View attachment 139490

Analysing all the images:-
View attachment 139491

Objects in To Confirm when default is not checked:-
View attachment 139492

Clarity of the wanted alert image is not the problem:-
View attachment 139493
Post a screenshot of your camera AI settings. The only two labels that ipcam-general have are person and vehicle.
 
Mike, I had no idea that ipcam-general is limited to only person and vehicle. I don't send emails or texts and would like AI triggers on near enough everything except spiders, webs, insects, rain, fog, swaying plants, dancing shadows and the like.

Screenshot 2022-09-10 203056.png
 
@Dave Lonsdale

Dave,

Sounds like what you want is ipcam-combined or you could download yolov5l which some have found to be more accurate (including myself) and has the ability to identify many different things.

Main Github page for YOLOv5:

Download link for the file you need to put in your code ai assets folder (C:\Program Files\CodeProject\AI\AnalysisLayer\CustomDetection\assets):
 
Last edited:
  • Like
Reactions: clk8