CodeProject.AI Version 2.0

MikeLud1

IPCT Contributor
Apr 5, 2017
2,278
4,371
Brooklyn, NY
CodeProject.AI version 2.0 was released Jan 16, 2023, this thread is for all topic CodeProject.AI version 2.0.


If you need any info on CodeProject.AI version 1.x please refer to the below link (there is some discussion on v2.0)
 
Last edited:
Thank you, will be following this just waiting to see all the bugs if any to be ironed out sounds like it’s almost there already.

Wanna make the switch from DS specifically since this also brings plate reading
 
  • Like
Reactions: tech101
Would it be worthwhile adding important info to the start of this thread maybe else it’ll be a case of the same questions being asked and the reply being to go and look back through the other thread? Just a thought.

You mean you don't want a post every other day asking what objects are in which models :lmao:

Or when Coral will be supported LOL
 
Last edited:
I have the problem with the custom models and service, too.
And on the other system where I have CPAI running on another host in docker it complaints on startup that it couldn't start AI (no, start/stop with BI is NOT checked) but a moment later reports tha AI is responding. And there I only get the folder-dialog when looking for custom models.
And I can't check the CUDA and ALPR-boxes, they're greyed out.
Seems there are some hiccups at the moment.
I'll wait a few days if that gets resolved, otherwise I'll report it to Ken. But could be possible that he's already aware of that, and has enough reports at the moment.
 
  • Like
Reactions: tech101
Awesome thanks for all the efforts.

I'm having the same issue others are in that BI doesn't detect custom models when started as a service. I tried the guidance here from @JNDATHP but whenever I start BI as a service the custom models field is blank:

I think few folks were having similar issue yesterday people who are running CPAI, seems to be in the process of being reported to the BlueIris Team.

 
This not seeing the models issue in BI does not affect CPAI at all. As long as you specify in each cam what model you want to use it works fine. All that really shows are which models are installed in CP. The selection is still in each camera's settings. Ken has already replied he is aware of this.

Screenshot 2023-01-24 191306.png

UPDATE: Issue fixed in 5.6.9.2
 
Last edited:
Testing and tuning -> Analyze with AI doesn't appear to be working for me. Actual clips are getting analyzed. Is it just me?
 
It is working for me.
Working for me again after upgrading to 5.6.9,2

I have two cameras with "ipcam-general" in the custom models field, but it is running the images through ALL the models. Is there a way to get it to only use the one I entered? Oddly, I have two other cameras that are set to "license-plate" and the AI analysis window shows that those are only being run through license-plate.

Edit: now I'm not sure what is going on. It looks like all of the cameras are running all of the models. Response times are really slow.

Edit:
I've also seen this several times:

1674672904083.png

I tried entering:
"license-plate;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;ipcam-general:0 "
for the custom models to disable the other ones. The AI analysis window still shows all the model being checked.

This camera is set for "to confirm" and "Mark as vehicle" = DayPlate,NIghtPlate,

1674673053498.png


yet this alert image seems to have been generated as soon as it detected "vehicle" or "car"

1674673109299.png

It should never have detected a car or vehicle, and should have only fired the alert image when it detected a DayPlate or NIghtPlate.

The custom models seem to be a mess in this version. I'm sure Ken will have it sorted soon.



Even More Edit - I re-read the documnetation, so I tried treating the list as an exclude list, entering everything I didn't want to use. So "objects:0;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;license-plate:0" to run just ipcam-general and objects:0;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;ipcam-general:0" to run just license plates. Looking at the AI analysis windows it still runs everything through all models... but it returns "Nothing found" even though the analysis window has LOTS of detected objects, including 'vehicle' which is in the 'to confirm' box.

This analysis returned "nothing found" when using those settings even though it clearly detected objects.

1674674253209.png
 
Last edited:
Working for me again after upgrading to 5.6.9,2

I have two cameras with "ipcam-general" in the custom models field, but it is running the images through ALL the models. Is there a way to get it to only use the one I entered? Oddly, I have two other cameras that are set to "license-plate" and the AI analysis window shows that those are only being run through license-plate.

Edit: now I'm not sure what is going on. It looks like all of the cameras are running all of the models. Response times are really slow.

Edit:
I've also seen this several times:

View attachment 151930

I tried entering:
"license-plate;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;ipcam-general:0 "
for the custom models to disable the other ones. The AI analysis window still shows all the model being checked.

This camera is set for "to confirm" and "Mark as vehicle" = DayPlate,NIghtPlate,

View attachment 151935


yet this alert image seems to have been generated as soon as it detected "vehicle" or "car"

View attachment 151936

It should never have detected a car or vehicle, and should have only fired the alert image when it detected a DayPlate or NIghtPlate.

The custom models seem to be a mess in this version. I'm sure Ken will have it sorted soon.



Even More Edit - I re-read the documnetation, so I tried treating the list as an exclude list, entering everything I didn't want to use. So "objects:0;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;license-plate:0" to run just ipcam-general and objects:0;actionnetv2:0;ipcam-animal:0;ipcam-combined:0;ipcam-dark:0;ipcam-general:0" to run just license plates. Looking at the AI analysis windows it still runs everything through all models... but it returns "Nothing found" even though the analysis window has LOTS of detected objects, including 'vehicle' which is in the 'to confirm' box.

This analysis returned "nothing found" when using those settings even though it clearly detected objects.

View attachment 151949
Use comma delimited for example: ipcam-combined, ipcam-dark do not use semicolon
 
Last edited:
  • Like
Reactions: Alan_F
Use comma delimited for example: ipcam-combined, ipcam-dark do not use semicolon

Tried:
objects:0,actionnetv2:0,ipcam-animal:0,ipcam-combined:0,ipcam-dark:0,ipcam-general:0

Check my work... that is a direct cut and paste from the UI.

Results in all models being used and 'nothing found' result even though the analysis window shows listed 'to confirm' objects from the non-excluded custom models (in this case, there is a DayPlate in the image, and that string above excludes everything BUT license-plate)
 
Also updated in 5.6.9.2: "use CUDA" in the general settings of BI now is "use GPU" to reflect the need to check that to get CPAI to use GPU in the .NET version.
 
Does anyone know how to get the JSON payload out to a webservice in an alert?

I tried the following but it sends nothing (presumed the required objects could be left blank to send everything?):

alert-ai.jpg