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 changed by RegEdit (as admin) but it reverts back after a reboot; 'Use custom model' still greyed out.
Did you modify both the deepstack_custom and deepstack_ custompath registry keys located in Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Perspective Software\Blue Iris\Options\AI as shown below?

Screen Shot 2022-07-30 at 10.37.28 PM.png
Screen Shot 2022-07-30 at 10.37.46 PM.png
 
  • Like
Reactions: misterfredsr
Hi guys, perhaps a dumb question. I upgraded today to CodeProject 1.5.5.1 (from 1.5, did the uninstall first, restarted, then installed 1.5.5.1).
Running the windows install.

On a restart my BlueIris does not detect 'person' anymore. It picks up 'car' just fine. What would cause that?

I haven't changed any settings but noticed that some of the modules are not green? If that could be the problem how can I turn them on?

Could someone point me in the right direction. Cheers

Screen Shot 2022-07-31 at 5.06.56 pm.png
 
Last edited:
If that could be the problem how can I turn them on?
It's not a problem unless you are trying to use custom models. I'm not sure why it's not recognising a person. Is the default object detection enabled? If so, you could try disabling it and use ipcam-general as a custom model.

On another note there is supposed to be a new release later this week that supports CUDA so this might change things a bit.

Screen Shot 2022-07-31 at 7.26.48 AM.png
 
  • Like
Reactions: sebastiantombs
CodeProject.AI Update

There should be a new release out next week with GPU support.

Since they are actively developing CodeProject.AI this is the time to request features to be added so if you have any ideas or features you would like to see head over the their forum and post your ideas or features you would like to see

I requested they add my custom models and add the ability to benchmark them which they did.

CodeProject.AI Forum

1659290145670.png
 
So the frame processing speed is essentially doubled with the GPU version. That is a really significant improvement.
 
It's not a problem unless you are trying to use custom models. I'm not sure why it's not recognising a person. Is the default object detection enabled? If so, you could try disabling it and use ipcam-general as a custom model.

On another note there is supposed to be a new release later this week that supports CUDA so this might change things a bit.

View attachment 135070

Yes, default object detection is on.

When I disable it, the Use custom model folder is still greyed out. How can I fix that, tried rebooting without luck.
 
When I disable it, the Use custom model folder is still greyed out.
Is there a check mark in the "Use custom model folder:" option? If so, custom models should work. If not, you'll have to modify the windows registry as previously mentioned in this thread.
 
Hi all, Blue Iris support advised me to try SenseAI after I flagged some missed obvious motion Deepstack events with them. I run only 4 cameras so the setup is rather straightforward - I'm after AI reliability rather than speed. Haven't delved into custom models yet... but will keep an eye on your SenseAI experiences in this forum.

I currently run Deepstack on "CPU only". For SenseAI, did I read correctly that "CPU only" may not be an option after GPU support is released? Surely it has to cater for the non power users?? I am planning to replace Deepstack with SenseAI this weekend... but if "CPU only" is not offered, I might have to wait?
 
I currently run Deepstack on "CPU only". For SenseAI, did I read correctly that "CPU only" may not be an option after GPU support is released? Surely it has to cater for the non power users?? I am planning to replace Deepstack with SenseAI this weekend... but if "CPU only" is not offered, I might have to wait?
CPU only is the foundation in which SenseAi was built so I’m pretty sure it’s not going away.
 
  • Like
Reactions: macthaidhg
CodeProject.AI Update

I am testing version 1.5.6 which has GPU support (CUDA), below are some details

1659756493321.png

Default object detection (80 objects) that uses an ONNX AI model works great. Using the Benchmark I am seeing about 16 msec detection times. Blue Iris is getting about 160 msec detection times for the first 2 detections then speeds up to about 35 msec detection times.

1659756384166.png
1659756201209.png

Custom object detection need some work. Using the Benchmark I am seeing about 60 msec detection times. Blue Iris is getting about 600 msec detection times

1659756696397.png
1659756893425.png
 
Last edited:
So, if I wanted to run it and have it be the equivalent of the default blue Iris/Deepstack, no custom models, no faces:
I would only need the box, in your first screenshot, 'Codeproject.AI Object Detection' to be selected?

I also want to take a look at the GPU times

Thank you