CodeProject.AI Version 2.0

[Craig G]

Something I can't quite explain yet.
Double alerts prompts - didn't used to happen

View attachment 162007

84% , 82%

Any ideas why?

I'm assuming it's something to do with how many images I'm feeding to AI within a set timeframe and it's seeing the same frame as multiples?

 

[Craig G]

for info. My BI settings for that indoor camera

 

[David L]

This is what i presently have, since I am always playing with this:

 

[David L]

Something I can't quite explain yet.
Double alerts prompts - didn't used to happen

View attachment 162007

84% , 82%

Any ideas why?

What does your Blue Iris Status Log show?

Mine above shows the stationary vehicles and the one that caused the alert.

 

[Craig G]

What does your Blue Iris Status Log show?

Mine above shows the stationary vehicles and the one that caused the alert.

Just the 82%

 

[David L]

Haha, beat this:



Just don't remember what I did to fix it, but it was around the time I switched to .NET and started playing with custom models.

 

[David L]

Also have been playing with Delivery model. This one is way kewl. I have Amazon, FedEx, UPS & USPS:

People can mask USPS AI detections

I implemented a custom DeepStack model to alert me when the mail is delivered because my mailbox is down the street from my home. I'm using a camera mounted near the street that also detects people and vehicles (including ALPR). The post truck detection works but will fail if the event is...

ipcamtalk.com

This is Clone:

 

[David L]

View attachment 162012

Just the 82%

What about your AI log


Open this window then click start the alert video and the AI info will show...

 

[Craig G]

I live in the UK. It won't work for me unless there's an AI preset for 'Fake palm leaves floor with spike pit underneath' to catch these fools rocking up on my drive

 

[Craig G]

What about your AI log
View attachment 162019

Open this window then click start the alert video and the AI info will show...

I guess I need to activate the analysis on all cameras to make that work so I record the dat files?

 

[David L]

I guess I need to activate the analysis on all cameras to make that work so I record the dat files?

So my thought was getting the AI info on your alert of the cat. The Alert you shared when you double click on it does it start the video of that event/alert? If so, then launch the Log window and go to AI, then go back to the Alerts and double click on the one with the 2 cats, this way you can see what AI did.

 

[Craig G]

I've enabled the AI analysis on that indoor cam. I'm sure they'll roam about in the night so I can compare tomorrow morning

 

[Craig G]

So my thought was getting the AI info on your alert of the cat. The Alert you shared when you double click on it does it start the video of that event/alert? If so, then launch the Log window and go to AI, then go back to the Alerts and double click on the one with the 2 cats, this way you can see what AI did.

Quick follow up.

It's displaying those two different percentage detections in BI because it's parsing all models since I don't have a specific model defined on that camera.

 

[David L]

Quick follow up.

It's displaying those two different percentage detections in BI because it's parsing all models since I don't have a specific model defined on that camera.

View attachment 162066

Makes sense. What I am testing now is Cloning my CAMs. One for Delivery model. One for Persons, and one for Vehicles, this one I may abandon since it makes no sense. Love how all my times are in-sync...


So maybe you should only use general and animal for your indoor CAM. I like all you do is separate them by a comma in Custom models:

 

[David L]

If you do Clone be sure and check this off...



Also don't worry about Cloning causing the Camera to over work, the above stops anymore streaming from the Camera for the Clones (Zero Bitrate):

 

[actran]

BI 5.7.4.2, Upgraded CP.AI from 2.0.8 to 2.1.8. Enabled ALPR and YOLO.NET only, and with GPU.

Before, I was getting <100ms prediction times. Now, it 500ms or worse.
Are the prediction time calculation incorrect in 2.1.8? (I did not change any AI configs in BI so the images sent to CP.AI are the same sizes as before, and leveraging mostly custom model ipcam-general. Not using default detection.)

I am watching Windows Task Manager and I can see spikes in GPU so that I know CP.AI is using it.
And yes, I did reboot Windows server several times just to be sure.

 

[David L]

BI 5.7.4.2, Upgraded CP.AI from 2.0.8 to 2.1.8. Enabled ALPR and YOLO.NET only, and with GPU.

Before, I was getting <100ms prediction times. Now, it 500ms or worse.
Are the prediction time calculation incorrect in 2.1.8? (I did not change any AI configs in BI so the images sent to CP.AI is same sizes as before, , leveraging mostly custom model ipcam-general. Not using default detection.)

I am watching Windows Task Manager and I can see spike in GPU so that is being leveraged...
And yes, I did reboot Windows server several times just to be sure.

View attachment 162074View attachment 162075View attachment 162076

Try YOLOv5 6.2

Seems many here are running it with their GPUs

 

[David L]

BI 5.7.4.2, Upgraded CP.AI from 2.0.8 to 2.1.8. Enabled ALPR and YOLO.NET only, and with GPU.

Before, I was getting <100ms prediction times. Now, it 500ms or worse.
Are the prediction time calculation incorrect in 2.1.8? (I did not change any AI configs in BI so the images sent to CP.AI are the same sizes as before, and leveraging mostly custom model ipcam-general. Not using default detection.)

I am watching Windows Task Manager and I can see spikes in GPU so that I know CP.AI is using it.
And yes, I did reboot Windows server several times just to be sure.

View attachment 162074View attachment 162075View attachment 162076

Noticed your Server is Offline. Do you know if this is common. Mine goes Offline too but I am presently working on an Internet Rule/Connection which may be the cause.

 

[actran]

@David L Yeah, I just noticed it's listed as "offline" in my screenshot. Reloaded my CP.AI and it's saying online now.

 

[MikeLud1]

BI 5.7.4.2, Upgraded CP.AI from 2.0.8 to 2.1.8. Enabled ALPR and YOLO.NET only, and with GPU.

Before, I was getting <100ms prediction times. Now, it 500ms or worse.
Are the prediction time calculation incorrect in 2.1.8? (I did not change any AI configs in BI so the images sent to CP.AI are the same sizes as before, and leveraging mostly custom model ipcam-general. Not using default detection.)

I am watching Windows Task Manager and I can see spikes in GPU so that I know CP.AI is using it.
And yes, I did reboot Windows server several times just to be sure.

View attachment 162074View attachment 162075View attachment 162076

What does the detection times show if you run a test using Explorer