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/

Apologies if this has been asked and answered previously, but I just switched over to CP AI-Server today and everything is working fine. However, if I open up the dashboard and look at the log, there is object detection happening every 2 minutes for each camera with AI enabled even when the cameras aren't triggered. In the screenshot below I disabled AI on every camera except for one and you can see I'm getting calls to ipcam-general and ipcam-animal (which are the custom models I have set up for that camera) every 2 minutes. I am running the GPU version of CP AI-Server in docker - version 1.6.8. My Blue Iris version is 5.6.4.2. Any idea what's going on here?

cpai.png
 
Apologies if this has been asked and answered previously, but I just switched over to CP AI-Server today and everything is working fine. However, if I open up the dashboard and look at the log, there is object detection happening every 2 minutes for each camera with AI enabled even when the cameras aren't triggered. In the screenshot below I disabled AI on every camera except for one and you can see I'm getting calls to ipcam-general and ipcam-animal (which are the custom models I have set up for that camera) every 2 minutes. I am running the GPU version of CP AI-Server in docker - version 1.6.8. My Blue Iris version is 5.6.4.2. Any idea what's going on here?

View attachment 146383
This is normal If you have Detect/ignore static objects enabled, what Blue Iris is doing is checking for static objects.

1669179130700.png
 
This is normal If you have Detect/ignore static objects enabled, what Blue Iris is doing is checking for static objects.

Waitaminute.

I always thought this setting (when enabled) IGNORES static objects.

You're telling me this setting (when enabled) tells it to LOOK FOR static objects??

How can the AI even know if an object is static or not if it only analyzes individual pictures and doesn't compare them? Doesn't BI determine whether there's movement or not via the trigger settings?
 
  • Like
Reactions: Dave Lonsdale
It can't ignore static objects unless it is checking when there is no motion for static objects and comparing to last image....otherwise what if your field of view goes 6 hours between motion at night - then a parked car will show up as non-static just based on how the headlight hits it differently....

BI motion has nothing to do with how AI determines if an object is static...
 
BI motion has nothing to do with how AI determines if an object is static...

So how does CP.AI determine if an object is static? And does the setting (when enabled) ignore static objects or not?
Am I right in thinking that if I disable the setting, it will stop detecting static objects and only focus on objects in motion?
 
It determines if an object is static by comparing the previous image. If you turn it off then a non moving object may get tagged as moving....Deepstack did the same thing BTW...read the BI help file...
 
@MikeLud1 Mike, I notice in your reply on static objects that your settings have "Use mainstream". I'm slowly catching on but I thought you always recommended to uncheck this option. Is that right?
 
@MikeLud1 Mike, I notice in your reply on static objects that your settings have "Use mainstream". I'm slowly catching on but I thought you always recommended to uncheck this option. Is that right?
The screenshot was from my LPR camera, mainstream is need for the ALPR module that I am working so when it crops a plate it is at full resolution.
 
After dark this afternoon a delivery driver wearing a Hi Vis jacket with lots reflective strips showed up well under my cameras IR lights, as though we was wearing an illuminated jacket. He was missed by all cameras in CP AI but did trigger clips in all cameras in BI.
Has the ipcam-dark model been trained on such an image?
 
After dark this afternoon a delivery driver wearing a Hi Vis jacket with lots reflective strips showed up well under my cameras IR lights, as though we was wearing an illuminated jacket. He was missed by all cameras in CP AI but did trigger clips in all cameras in BI.
Has the ipcam-dark model been trained on such an image?
Not sure if it’s trained for that but occasionally I get missed delivery men as well using ipcam-general. It’s not often however. I’d put it at a 99% success rate which is good. No AI will be 100%.
 
  • Like
Reactions: sebastiantombs
Just to add some clarification, in Testing and Tuning CP AI didn't draw any box around the delivery driver, it did around his truck by the road and my parked vehicle in our drive. BI though did have him covered in small colored squares so appeared in the list of All clips. If I'd set the lowest of confidence levels in AI I expect that he still wouldn't have shown up in my list of Alert clips.

I suppose it's also impressive how modern reflective clothing is so effective, I could see him but not AI.
 
Just to add some clarification, in Testing and Tuning CP AI didn't draw any box around the delivery driver, it did around his truck by the road and my parked vehicle in our drive. BI though did have him covered in small colored squares so appeared in the list of All clips. If I'd set the lowest of confidence levels in AI I expect that he still wouldn't have shown up in my list of Alert clips.

I suppose it's also impressive how modern reflective clothing is so effective, I could see him but not AI.
Cloaking device:rofl: