CodeProject.AI Version 2.5

I have been running 2.6.2. When I installed it I only used Yolo8, no other modules installed, and it had been like that for the first two weeks of my checking it, just Yolo 8 installed. Then suddenly I looked yesterday and it had installed Face, Yolo.net, and Yolo 5.6.2.
How does that happen?
 
My system becomes unresponsive every 4 to 5 days, and the only way I can get it back is to do a rude shutdown (i.e. cut the power) and reboot. Could be BI. CPAI, windows, or maybe even a hardware issue. I don't have any time to dig deeper, so I just live with it. No point in me complaining if I don't spend the time to isolate what's happening.
 
My system becomes unresponsive every 4 to 5 days, and the only way I can get it back is to do a rude shutdown (i.e. cut the power) and reboot. Could be BI. CPAI, windows, or maybe even a hardware issue. I don't have any time to dig deeper, so I just live with it. No point in me complaining if I don't spend the time to isolate what's happening.
Make sure you have Auto start/stop with Blue Iris unchecked.

1716052952419.png
 
Im still working with an old version of BI, with latest version of AIToo,l and 1st GPU version of Deepstack, which for me, is rock solid reliable.... (2nd Deepstack GPU version has a memory leak.) I figured enough time passed to work out all the issues with CPAI so I started testing last week.

Is there any way to run multiple instances of CPAI in or out of Blue Iris?
I see in the screenshots of BI its has a box for instances which I don't have since I'm still using an old version of BI pre Deepstack integration.

I run 4 instances of Deepstack for 10 cams with a Nvidia P400 and my average queue time is 30ms with AITool.

I couldn't figure out how to run multiple instances of CPAI and within a few hours queue had built up over 100 images and was unusable.

What am I missing?
 
I am currently running both yolov562 and yolov8. Using v8 to detect people and vehicles results in far less fals-positives but tends to be a bit slower (better then fals alerts in my case). Still using v562 only at my “pond camera” for detecting herons so the alarm gets triggered. Does anyone know if it is possible to use the ipcam-animal models in v8 so I can discontinue v562 and free up some resources? Maybe this will be available in the next update?
 
I updated BI yesterday, and now, after a rebuild and a system restart, seems to be working fine with AI captures and alerts saved with my older version of CP AI (2.5.1). Since the last attempt to update CP AI a few months ago resulted in needing to go back to 2.5.1, I wanted to get a community read on the stability of 2.6.5 now before giving the upgrade an attempt, and to see if there are best practice steps to take during the process, like stopping, deleting previous or deleting old alert files/folders.

Note: no NVDIA GPU used, just Intel GPU. Using only YOLOv5.NET
 
I updated BI yesterday, and now, after a rebuild and a system restart, seems to be working fine with AI captures and alerts saved with my older version of CP AI (2.5.1). Since the last attempt to update CP AI a few months ago resulted in needing to go back to 2.5.1, I wanted to get a community read on the stability of 2.6.5 now before giving the upgrade an attempt, and to see if there are best practice steps to take during the process, like stopping, deleting previous or deleting old alert files/folders.

Note: no NVDIA GPU used, just Intel GPU. Using only YOLOv5.NET
I'm using BI 5.9.2.3 and CPAI 2.6.5, no GPU (intel built in), ipcam-combined, and YOLOv5.net also. All running ok and doing what is expected. (my update to BI 5.9.2.3, from 5.9.2.1) was ok but at first the CPU was running very high...like 90%, for about 30 minutes just after the update, but then settled back to its usual 10-15%, so I think it was just settling in or something. Otherwise, all good.
 
  • Like
Reactions: bug99
Nope.

it downloaded fine. I unpacked the zip file and ran the exe, with no errors. Windows shows CodeProject.AI Server running, but the CodeProject AI dashboard shows it as offline and still at 2.5.1. I dont see an obvious way to kick it along, as it looks to be done doing whatever it was doing. I restarted the windows service (CPAI), but that did not do anything useful. Isn't there a simple stop, install and start sequence outside of the install exe for CP AI (and B)I?

Update: As a last ditch effort, i restarted the computer, and oddly the server started in the dashboard correctly and at the latest version, and BI is now using it fine as you please. I have no ide why.
 
Last edited:
Nope.

it downloaded fine. I unpacked the zip file and ran the exe, with no errors. Windows shows CodeProject.AI Server running, but the CodeProject AI dashboard shows it as offline and still at 2.5.1. I dont see an obvious way to kick it along, as it looks to be done doing whatever it was doing. I restarted the windows service (CPAI), but that did not do anything useful. Isn't there a simple stop, install and start sequence outside of the install exe for CP AI (and B)I?

Update: As a last ditch effort, i restarted the computer, and oddly the server started in the dashboard correctly and at the latest version, and BI is now using it fine as you please. I have no ide why.
Often I find you need to restart the service in windows, and also a simple refresh of the browser will show updates that otherwise don't seem to have taken. (reboot does both of course) Glad it's working...
 
  • Like
Reactions: David L
@MikeLud1 - have you upgraded to 2.6.5? For some reason individual CPAI models aren't starting at system boot. Under info I see AutoStart set to true, but when I boot up the Object Detection YOLO .NET module is ready but not running. ALPR seems to start fine at boot. Any thoughts?

NVM - I manually went in and edited the Object Detection .NET configuration file and changed AutoStart to "true" and all seems okay now. Not sure why it didn't install that way.
 
Last edited:
@MikeLud1 - have you upgraded to 2.6.5? For some reason individual CPAI models aren't starting at system boot. Under info I see AutoStart set to true, but when I boot up the Object Detection YOLO .NET module is ready but not running. ALPR seems to start fine at boot. Any thoughts?

NVM - I manually went in and edited the Object Detection .NET configuration file and changed AutoStart to "true" and all seems okay now. Not sure why it didn't install that way.
I was just going to send the configuration file with AutoStart changed to "true". There is a bug that is not saving any setting from the default settings.
 
  • Like
Reactions: David L
I for the life of me can not get license plate reader to work with CUDA and it starts and stops with CPU,

1718821142204.png1718821116133.png
 
I'm getting the following when trying to install the coral module, I have checked and my numpy is version 1.21.6. Python version is 3.9.13. Any ideas?

1718834439034.png