CodeProject.AI Version 2.0

[dsteve]

Is there any updated timeline on Coral TPU support?

Here's what they said a few days ago:

CodeProject.AI Server: AI the easy way.

Version 2.5.4. Our fast, free, self-hosted Artificial Intelligence Server for any platform, any language

www.codeproject.com

 

[sp3ctre]

What's the best settings? Can anyone share that here. I'm mainly talking about in the camera AI section for what I need to place. For reference I'm running an i5-6500 with the igpu

Also, does my lpr camera have to be recording to work with lpr?

I have the same CPU with the Intel 530 graphics.... should GPU be set to on or off for this chip?

 

[Tinman]

I have the same CPU with the Intel 530 graphics.... should GPU be set to on or off for this chip?

I myself would not use the 530 in GPU mode. It really struggled on mine and I have much better results using the CPU.

 

[truglo]

Hmm I thought GPU was just for cuda, not intel... but idk. Before I had a gpu I ran cpai with cpu, and it was decent (not too hard on the cpu, make times ~3x longer than with cuda... 200msec which is OK for most small/medium homes I think).

 

[Vettester]

So I've spent the last few days dialing in the AI settings for BI (5.6.9.2) and CPAI (2.0.6) and everything seems to be working well so I thought I would share my settings. First let me explain how my system is configured. I'm currently running 17 cameras (plus 3 cloned cameras) on an i7-4790, CPAI is enabled on all my cameras. The cameras are on various schedules so not all of them are running all the time, but typically there are 10-12 cameras enabled. My main BI computer supports intel GPU, but does not have an additional NVIDIA graphics card. I also have an i7-6700T that I have CPAI running on to help balance the CPU load. I have CPAI configured with GPU enabled on my main BI machine to handle only the motion detection from my LPR camera. The motion detection from all my other cameras is handled by the other PC running CPAI in CPU only mode.

Here's some screenshots of my configurations.



Note: BI and CPAI for my LPR camera are on 192.168.86.58 and all the other cameras are using CPAI on 192.168.86.57

And the results

 

[Vettester]

Updating to 5.6.9.3 causes AI to stop working.

 

[MikeLud1]

Updating to 5.6.9.3 causes AI to stop working.

View attachment 152048

Do the usual full reboot.

 

[Vettester]

Do the usual full reboot.

Reboot didn't help. Rolled back to 5.6.9.2.

 

[Tinman]

Updating to 5.6.9.3 causes AI to stop working.

View attachment 152048

Same here: I also noticed the custom models was blank again.

UPDATE: did a full reboot and all is ok...when will I learn always reboot on every update !

 

[gwithers]

BI 5.6.9.3 AI stills works normally on my setup. Still finding objects with the same expect times (~150-200 ms). Not sure what is going on but it apparently isn't just 5.6.9.3 itself but related to BI/CPAI installation topography in some fashion as well. I'm not running any CPAI locally on my BI instance, so I don't have any of the "auto start/stop", stuff selected either. Here is what I have that is working. I run two CPAI instances in docker containers on two VM's.

 

[MikeLud1]

Same here: I also noticed the custom models was blank again.

View attachment 152052

Check the main AI tab if models are not listed stop and restart CP.AI

 

[MikeLud1]

CodeProject.AI update, I had a call to day with Chris from CP and they were going to release an update today or very soon. This update will have v2.0 of my ALPR.

 

[harleyl7]

So I've spent the last few days dialing in the AI settings for BI (5.6.9.2), and CPAI (2.0.6) and everything seems to be working well so I thought I would share my settings. First let me explain how my system is configured. I'm currently running 17 cameras (plus 3 cloned cameras) on an i7-4790, CPAI is enabled on all my cameras. The cameras are on various schedules so not all of them are running all the time, but typically there are 10-12 cameras enabled. My main BI computer supports intel GPU, but does not have an additional NVIDIA graphics card. I also have an i7-6700T that I have CPAI running on to help balance the CPU load. I have CPAI configured with GPU enabled on my main BI machine to handle only the motion detection from my LPR camera. The motion detection from all my other cameras is handled by the other PC running CPAI in CPU only mode.

Here's some screenshots of my configurations.

View attachment 152028 View attachment 152029 View attachment 152033 View attachment 152030 View attachment 152035

Note: BI and CPAI for my LPR camera are on 192.168.86.58 and all the other cameras are using CPAI on 192.168.86.57

And the results

View attachment 152031

Thanks for this! I just put my 5 cameras on my CPAI running in docker on my unraid server and left my main BI/CPAI for my LPR. We will see how it goes.

 

[MikeLud1]

I just started a new thread for all topic CodeProject.AI ALPR below is the link

Blue Iris and CodeProject.AI ALPR

CodeProject.AI version 2.0 was released Jan 16, 2023 with my ALPR module, this thread is for all topic CodeProject.AI ALPR. To setup the ALPR you need one of the Object Detection module's and License Plate Reader module installed, You need to be on the current version of Blue Iris, below are my...

ipcamtalk.com

 

[wittaj]

So I decided to test out CodeProject on my non BI server, so all I did was download the latest CodeProject file and installed it. It is not connected to BI.

Initially it works and I put in the image and it detects the plate, etc. and then after a couple of minutes it is throwing me constant "server connection error is the server URL correct"

It is at

http://localhost:32168

and I also tried 127.0.0.1:32168 and rebooted and same issue?

But on the dashboard it shows the server is active?

 

[ohlin5]

Is there any way to get the yolov5l model working again? I don't know why, but the built-in models (I've tried ipcam-combined and ipcam-general) constantly recognize my dogs as people with 70-75+% confidence whereas prior to upgrade the yolov5l model worked really well. But after updating to BI 5.6.9.2/3 having the yolov5l.pt file in C:\Program Files\CodeProject\AI\AnalysisLayer\ObjectDetectionYolo\custom-models causes the BI "Use custom models:" box to just be blank. No amount of stopping/restarting/etc. seems to fix it. Even if I remove all other files from the custom-models directory it still refuses to load it. The only models I can get loaded are the built-in ones, when they're the ONLY files in that directory. What am I missing here?

 

[Tinman]

So I decided to test out CodeProject on my non BI server, so all I did was download the latest CodeProject file and installed it. It is not connected to BI.

Initially it works and I put in the image and it detects the plate, etc. and then after a couple of minutes it is throwing me constant "server connection error is the server URL correct"

It is at

http://localhost:32168

and I also tried 127.0.0.1:32168 and rebooted and same issue?

But on the dashboard it shows the server is active?

Did you put in an address in the dashboard of one of your cameras ? And the set that cam to use that AI in it's settings? Look above how Vettester does his remote AI.

 

[MikeLud1]

Is there any way to get the yolov5l model working again? I don't know why, but the built-in models (I've tried ipcam-combined and ipcam-general) constantly recognize my dogs as people with 70-75+% confidence whereas prior to upgrade the yolov5l model worked really well. But after updating to BI 5.6.9.2/3 having the yolov5l.pt file in C:\Program Files\CodeProject\AI\AnalysisLayer\ObjectDetectionYolo\custom-models causes the BI "Use custom models:" box to just be blank. No amount of stopping/restarting/etc. seems to fix it. Even if I remove all other files from the custom-models directory it still refuses to load it. The only models I can get loaded are the built-in ones, when they're the ONLY files in that directory. What am I missing here?

You can enable the default model to use the large yolov5 model

 

[wittaj]

Did you put in an address in the dashboard of one of your cameras ? And the set that cam to use that AI in it's settings? Look above how Vettester does his remote AI.

I don't have any cameras connected to it at the moment, nor is BI connected to it. I am simply testing CodeProject by itself with no connection to anything, so different than Vettster as he has BI on one computer and AI on another. That is not what I am doing at the moment.

I am just trying out the upload image to use to test/detect feature of things or plates.

It works for the first minute or so after I install it and then after that is keeps throwing that error in the log and says server timed out, even though it shows it is active and the response times before it went wonky were fine.

I am trying this feature locally (taken from one of Mike's posts, but using my photo instead) and mine shows API server online just like this, but in the log it keeps throwing that error):

 

[ohlin5]

You can enable the default model to use the large yolov5 model
View attachment 152108

Ok...if I do that do I need to mess with the files in C:\Program Files\CodeProject\AI\AnalysisLayer\ObjectDetectionYolo\custom-models at all? Previously if I remember correctly I had to add the yolov5l.pt file here and then specify that custom model in each of my camera's AI trigger settings box...but post upgrade that yolov5l.pt file is no longer present in C:\Program Files\CodeProject\AI\AnalysisLayer\ObjectDetectionYolo\custom-models.