Blue Iris "Out of memory" with 40GB unused

R

randylight

Young grasshopper
Jan 10, 2022
34
16
Michigan
I recently noticed some situations in which it was very obvious that there was a person in the frame and BI does not even get triggered. When I go back and analyze the video with the AI it sees it perfectly. I started looking at the AI analysis details and for the situations where it was an obvious miss it seems that the file sizes for the AI details get bigger and bigger until finally, something happens and everything goes back to normal.

When I look at the file sizes for AI details on a normal trigger they're anywhere from about 200 - 500 KB. When I go to the AI details on a trigger that should have sent an alert the file size is nearly 25 MB and when I try to open the details in BI it gives me a "out of memory" error message.

Does anyone know why this may happen and how to prevent this? My system still has nearly 40GB of unused memory so I don't see that as an issue.
 

Attachments

  • Out of memory.png
    Out of memory.png
    7.9 KB · Views: 17
  • 25MB.png
    25MB.png
    35.4 KB · Views: 17
Review your logs for additional points of data that may be actionable.
 
  • Like
Reactions: looney2ns
Nothing in my logs is abnormal. At 11:21 I have a .dat file that's 45 MB for "Garage" and when I try to open that file in BI it gives me that "out of memory" message. It manages to send an alert on this one but a lot of the time it will not recognize any objects.
 

Attachments

  • logs.png
    logs.png
    36.8 KB · Views: 8
I may be the wrong person to help you, so sorry. I would need to know which AI you are using and the settings that you have implemented.

But there are many express here that I am sure will chime in and help out
 
I recently switched to CodeProject AI but this has been happening since before that. It’s only on 1 camera so I’m hoping it’s not an issue with that camera. I just updated the firmware as well and that didn’t help the situation.

I increased the buffer from 6 MB to 20 MB but I’m not exactly sure what that’s for and I'm out of options so I’m throwing shots in the dark.
 
I found some interesting information on why these file sizes are so large. I was able to open one of these larger .dat files and noticed something odd. The issue seems to be with the "static" analysis, or whatever this is. When I drill into the details on the IPcam-combined portion all I see is a single "nothing found" message. However, when I drill into the "static" option it shows THOUSANDS of the same "car 45%" or whatever. There are thousands of these on each of the additional images.

This has to be the issue but I'm not sure how to solve it. Most of the time when I open these up it crashes BI.
 

Attachments

  • static.png
    static.png
    727.1 KB · Views: 17
  • static_2.png
    static_2.png
    38.7 KB · Views: 17
Can you post the BI/AI settings and the camera's AI settings ? The CodeProject Dashboard/status screen shot would be helpful as well.
 
It sounds like the periodic check to look for stationary items is being retained/stored in that file. Turn off the stationary object on that camera and see if it goes away.
 
Here are my settings for BI AI and CodeProject. I did turn off the "detect static" option for now and nothing has errored out but there isn't much activity yet today.
 

Attachments

  • CodeProject.png
    CodeProject.png
    127 KB · Views: 21
  • BI-AI 2.png
    BI-AI 2.png
    67.9 KB · Views: 22
  • BI-AI-Garage.png
    BI-AI-Garage.png
    66.7 KB · Views: 22
Just try changing the cameras AI settings to this and see if it helps your issue. You do not need the objects:0 in the setting. If it helps, then change one thing at a time back to see if it works.

Screenshot 2022-12-19 101438.png
 
  • Like
Reactions: looney2ns
I remember for some reason someone suggested that I have that "objects:0" in there so I removed that and it has been more accurate. That objects model seems to be more accurate than the custom combined model. I'm still getting a lot of static images on each imaged analyzed on every trigger. None of these are even close to the images it's saying it's detecting.

One of my .dat files was over 100 MB so i would like to find out why that's possible to have so many static images on a single trigger. I sent an email to BI to hopefully get more insight on that.
 

Attachments

  • static_3.png
    static_3.png
    1.2 MB · Views: 13
randylight said:
I remember for some reason someone suggested that I have that "objects:0" in there so I removed that and it has been more accurate. That objects model seems to be more accurate than the custom combined model. I'm still getting a lot of static images on each imaged analyzed on every trigger. None of these are even close to the images it's saying it's detecting.

One of my .dat files was over 100 MB so i would like to find out why that's possible to have so many static images on a single trigger. I sent an email to BI to hopefully get more insight on that.
Click to expand...
None of my custom models have boat in them, the default objects model is detecting them.
 
  • Like
Reactions: fenderman
MikeLud1 said:
None of my custom models have boat in them, the default objects model is detecting them.
Click to expand...
This issue was present way before I turned off the default objects model. Even with it on, would it be normal to have over 250 static objects detected that aren't actually there? I turned back on the "detect static" option while the amazon driver came up to the house and on all 3 cameras there were hundreds of false static objects detected. Something isn't right.
 
You must log in or register to reply here.
Top Bottom