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/

The direction of travel with having AI working with Blue Iris seems less to be DeepStack more Sense AI, so I am up for moving from DeepStack to Code Project AI Service.
However-
Does Sense AI still depend on having DeepStack installed? Can I uninstall DeepStack completely yet? Here I am on CodeProject.AI.Server-1.5.6.2.
Why is in BI global settings AI "Use custom model folder" still greyed out despite my regedit? Several folk have asked this question which I take to be mostly unanswered.
I get no AI unless I tick the box "Default object detection", but then it finds I have many potted plants and the rear of my car is identifies as a toilet. Is this DeepStack still at work?
Bring on use of a Custom model, but it's not working here yet.

I know we all need to keep working at it to get the result we are all searching for, good AI in Blue Iris.
 
toastie said:
The direction of travel with having AI working with Blue Iris seems less to be DeepStack more Sense AI, so I am up for moving from DeepStack to Code Project AI Service.
However-
Does Sense AI still depend on having DeepStack installed? Can I uninstall DeepStack completely yet? Here I am on CodeProject.AI.Server-1.5.6.2.
Why is in BI global settings AI "Use custom model folder" still greyed out despite my regedit? Several folk have asked this question which I take to be mostly unanswered.
I get no AI unless I tick the box "Default object detection", but then it finds I have many potted plants and the rear of my car is identifies as a toilet. Is this DeepStack still at work?
Bring on use of a Custom model, but it's not working here yet.

I know we all need to keep working at it to get the result we are all searching for, good AI in Blue Iris.
Click to expand...
You do not need DeepStack installed for CodeProject.AI to work. If you are using a Nvidia GPU you still need the current versions of CUDA and cuDNN
Post a screenshot of your main IA settings. also my custom model names has changed to ipcam-general from general for example
Also Ken (Blue Iris) still needs to work on the integration.
 
  • Like
Reactions: looney2ns
Hmm... odd!

I am running 1.5.6.2 and latest 5.6.03 (BI) with GPU support... i've tried to reboot multiple times etc.. but for some reason, no matter what I do, I keep getting "Nothing found" on alerts... even when I playback with AI overlays on, nothing gets picked up there either...

Note: it was running fine on 1.5.6.0...

Any ideas?

1660974858874.png

1660974891711.png
1660975048119.png
 
joshwah said:
Hmm... odd!

I am running 1.5.6.2 and latest 5.6.03 (BI) with GPU support... i've tried to reboot multiple times etc.. but for some reason, no matter what I do, I keep getting "Nothing found" on alerts... even when I playback with AI overlays on, nothing gets picked up there either...

Note: it was running fine on 1.5.6.0...

Any ideas?

View attachment 137226

View attachment 137227
View attachment 137228
Click to expand...
What version of CUDA and cuDNN do you have installed? Also what GPU do you have?
From PowerShell run as administrator execute "nvidia-smi" and post a screenshot of the results
Try changing localhost to "127.0.0.1"

1660975687074.png
 
toastie said:
Why is in BI global settings AI "Use custom model folder" still greyed out despite my regedit? Several folk have asked this question which I take to be mostly unanswered.
Click to expand...

Same here. I worked for about 4 hours trying to get this working, reading and re-reading this thread which I have been following. Only default worked for me after uninstalling Deepstack, installing, uninstalling, reinstalling (which takes at least 20 minutes on my pc), multiple reboots. I finally uninstalled it, reinstalled Deepstack which took about 5 minutes and called it a day. I'll try again at some point but I got exhausted, maybe it will go smoother next time.
 
  • Like
  • Wow
Reactions: wantafastz28, looney2ns and sebastiantombs
This is hard to understand why it works so easy for some and not for others. So I am posting these pics to show my settings that is working perfect. Keep in mind this is a CPU install on BI 5.6.0.2 W11 I7-6700k system.

1. I did do a clean un-install of Deepstack long ago, which anything left in the temp folders are gone as well.
2. Have BI running
3. Install CodeProject
4.Restart pc

Thats it nothing was modded in my registry or anything, all stock. Keep in mind, you can change the settings later to your liking, but this will produce results.

Screenshot 2022-08-20 081407.pngScreenshot 2022-08-20 081443.pngScreenshot 2022-08-20 081600.png
 
  • Like
Reactions: gwminor48
Tinman said:
This is hard to understand why it works so easy for some and not for others.
Click to expand...
When I updated to 1.5.6.2 the install didn't go so well. Thinking I had learned from a previous failed update I uninstalled SenseAI and started with a fresh install. The installation took about 40 minutes and when it was finished it didn't say anything about rebooting so I tried opening the web UI and it couldn't find the server. I rebooted the computer and open the web UI again. This time it found the server but was not processing anything so I ran the install again. This time I picked the repair installation option and the installation processed started again. After about 20 minutes the repair process finished with a message stating a reboot was required. I rebooted again and everything started working.

As far as the CPU processing times I am seeing an improvement, but nowhere near the times you are reporting. However, I am running an i7-4790 so this is to be expected.

Screen Shot 2022-08-20 at 10.52.29 AM.png
 
  • Like
  • Wow
Reactions: TBurt, gwminor48 and sebastiantombs
  • Like
Reactions: looney2ns and gwminor48
Trying to do too much all at once. I hope he gives himself some breathing room with this. Don't want him to burn out trying to fix everything at once.

I do have to wonder if that isn't a factor, too, in why some are having problems when others aren't.
 
  • Like
Reactions: looney2ns and gwminor48
Tinman said:
This is hard to understand why it works so easy for some and not for others. So I am posting these pics to show my settings that is working perfect. Keep in mind this is a CPU install on BI 5.6.0.2 W11 I7-6700k system.

1. I did do a clean un-install of Deepstack long ago, which anything left in the temp folders are gone as well.
2. Have BI running
3. Install CodeProject
4.Restart pc

Thats it nothing was modded in my registry or anything, all stock. Keep in mind, you can change the settings later to your liking, but this will produce results.

View attachment 137259View attachment 137260View attachment 137261
Click to expand...

I notice your "Use custom model folder" is grayed out too, I guess this is normal? You choose model from AI dashboard? Will it let you choose multiple models from there?
 
gwminor48 said:
I notice your "Use custom model folder" is grayed out too, I guess this is normal?
Click to expand...
It is normal to have this grayed out. Custom models are located in the are located in the "C:\Program Files\CodeProject\AI\AnalysisLayer\CustomDetection\assets" directory by default. You enable custom models just like you do in deepstacks, just keep in mind the naming convention of the models have changed.

Screen Shot 2022-08-20 at 2.12.03 PM.png

You can also add and use additional models by putting them in the same directory. To simply things I have removed all the custom models that I'm not using from the directory. Not sure if this speeds things up at all, but I know when I analyze dat files it only shows the models I'm actually using.
 
Last edited:
  • Like
Reactions: slidermike, gwminor48 and sebastiantombs
After dealing with BI 5.6.0.4 issues that took hours to get my system stable. I decided to make the switch from DeepStack to CodeProject.AI on my main system. So far CodeProject.AI GPU (CUDA) is working with no issues.
I disabled all the modules except the Custom Object Detection Module. DeepStack was running three models and using more memory and loaded more processes then CodeProject.AI, see below.

1661044613297.jpeg

DeepStack memory usage
1661052763638.jpeg

CodeProject.AI memory usage
1661052798437.jpeg

DeepStack processes
1661044817785.jpeg

CodeProject.AI processes
1661044921393.jpeg
 
Last edited:
  • Like
Reactions: looney2ns, gwminor48, 105437 and 5 others
Great info @MikeLud1. I'm about ready to try CodeProject.AI. I plan on uninstalling Deepstack and then updating my CUDA from 10.5 to the required version before installing CodeProject.AI. Anything else I need to do? Thanks
 
You must log in or register to reply here.
Top Bottom