[tool] [tutorial] Free AI Person Detection for Blue Iris

Whats the main difference between using the AItool and deepstack directly in BI?

I just found a config guide today and setup my BI to use deepstack... then found this tool later! Just wondering if its worth undoing all my deepstack work and setting this up instead. (sorry for the super-newbie question - new Blue iris user here!)
 
Whats the main difference between using the AItool and deepstack directly in BI?

I just found a config guide today and setup my BI to use deepstack... then found this tool later! Just wondering if its worth undoing all my deepstack work and setting this up instead. (sorry for the super-newbie question - new Blue iris user here!)
I started with AITool and then started using the integrated DS in Blue Iris once it was released. For my situation, I haven't lost any functionality.
 
I started with AITool and then started using the integrated DS in Blue Iris once it was released. For my situation, I haven't lost any functionality.
Sorry but what do you mean by integrated deepstack in Blue Iris? I just bought this Blue Iris software and it seems quite nice. I just want to learn more. Thanks!
 
I think I got the point now. I was installing AI tool at the same time and started to receive those person, car etc. alerts on BlueIris. I tought that it was because of Aitool, but I was because I have Deepstack configured in global settings and enabled in trigger.

How evert what I do not understand that how can I get face rigorization working? I can add faces in pics but how should I see when some face is recognized?
 
I think I got the point now. I was installing AI tool at the same time and started to receive those person, car etc. alerts on BlueIris. I tought that it was because of Aitool, but I was because I have Deepstack configured in global settings and enabled in trigger.

How evert what I do not understand that how can I get face rigorization working? I can add faces in pics but how should I see when some face is recognized?
I don't use facial recognition since I really don't have a use case for it. Perhaps someone else will respond that does. Sorry I couldn't help.
 
I've been noticing high CPU utilization, went into my headless machine running BI and AI Tool is not responding. Functionally it is working as normal, the GUi just isn't usable and is glitchy. I also have a weird black box on the screen. Anybody else seeing this suddenly? I didn't update anything that I know of, it just started happening. Using 2.0.1205.7861, Windows GPU deepstack.
 
This forum does not presently appear very active so this information maybe redundant.
Myself and others have found a problem with recent releases of BI accepting Alerts from AI Tool (doesn't matter which version of AI TOOL you are using) .
I have gone back through recent releases of BI and found that version 5.5.0.15 is the last release that works with AI Tool.
If your BI software auto updates you may well have already come across this issue and decided to simply use the built in AI of BI.
Unfortunately unless your system is recording continuously BI will miss some alerts!
 
My AI tool (service) has recently been using high amounts of memory an CPU while just idling. If I shut the service down and try to open the GUI it freezes every time now.
Have to check if there is a new version.

Could it be related to recent update in windows 11?
What is the latest version of AI? Can;t even get to the update page on the tool GUI as it freezes.
 
Is there any way to have the "Prediction Tolerances" either be on a per Relevant Object basis, or allow a specific Relevant Object to disregard the specific Prediction Tolerance like we can for "Ignore Image Mask"?

For example, I use the "Limit Size in Pixels" width and height on a certain camera to avoid detecting things off in the background... but one object I am trying to detect is smaller than my min width even when it is in the area I want to detect. (Vehicle vs USPS logo)

If there were just a box on the USPS relevant object to ignore the limit size for that object only, that would do the trick.


Edit: Looks like this can be accomplished by setting up another camera in AI Tool and pointing it at the same BI Camera name :thumb:
 
Last edited:
Hey @Village Guy! I have not been on here in sometime glad to see this is still "kind of" active. I have had the "only 1 out of 2 'python.exe' processes did not start within30067ms" popping up for awhile now, truth be told it has been so long I do not remember when it started- BI/ AI are all working, I get alerts etc. but obviously there is some kind of issue- any clue what it is and how to correct?
Can't seem to post the log as a logfile? Here it is in a word -
 

Attachments

Hey @Village Guy! I have not been on here in sometime glad to see this is still "kind of" active. I have had the "only 1 out of 2 'python.exe' processes did not start within30067ms" popping up for awhile now, truth be told it has been so long I do not remember when it started- BI/ AI are all working, I get alerts etc. but obviously there is some kind of issue- any clue what it is and how to correct?
Can't seem to post the log as a logfile? Here it is in a word -
It started when you updated to DeepStack 2022.01.1 . The new version of DS only uses one python processes and AITool is looking for two.

Maybe it was just my system was jacked up from power outages and and not shutting down correctly, but the new GPU version of DeepStack 2022.01.1 is slow AF... The other day the wind was blowing hard and Deepstack couldn't keep up with all the jpegs BI was sending it. It build up 100 images in queued and ended up crashing AITool. I reverted back to DeepStack 2021.09.1 and as a bonus now I don't get the "only 1 out of 2 'python.exe' processes did not start within30067ms" error message.
 
Last edited:
  • Like
Reactions: Village Guy
It started when you updated to DeepStack. The new version of DS only uses one python processes and AITool is looking for two.

Maybe it was just my system was jacked up from power outages and and not shutting down correctly, but the new GPU version of DeepStack 2022.01.1 is slow AF... The other day the wind was blowing hard and Deepstack couldn't keep up with all the jpegs BI was sending it and it ended up crashing AITool. I reverted back to DeepStack 2021.09.1 plus now I don't get the "only 1 out of 2 'python.exe' processes did not start within30067ms"
Thanks. Anyway to make AI only look for 1 or is this something VorlonCD has to recode/fix?
 
Hey @Village Guy! I have not been on here in sometime glad to see this is still "kind of" active. I have had the "only 1 out of 2 'python.exe' processes did not start within30067ms" popping up for awhile now, truth be told it has been so long I do not remember when it started- BI/ AI are all working, I get alerts etc. but obviously there is some kind of issue- any clue what it is and how to correct?
Can't seem to post the log as a logfile? Here it is in a word -
I experienced the very same problem two weeks ago setting up a new system and never got to the bottom of the issue. The post from @whoami ™ is enlightening, thank you for that info.
I guess you have two choices, install an older version of DeepStack or use the builtin DeepStack support of BI.
I finally gave up fighting the tide last week and removed aitool from my system in favour of the built in BI recognition capability. It does not work very well with recording only on alert so I now record continuously and so far I'm very happy with the result.
 
I experienced the very same problem two weeks ago setting up a new system and never got to the bottom of the issue. The post from @whoami ™ is enlightening, thank you for that info.
I guess you have two choices, install an older version of DeepStack or use the builtin DeepStack support of BI.
I finally gave up fighting the tide last week and removed aitool from my system in favour of the built in BI recognition capability. It does not work very well with recording only on alert so I now record continuously and so far I'm very happy with the result.
Curious, is the BI integrated setup able to send a Pushover notifaction if a person is detected that is bigger than 2% of the image between 23:00:00-05:00:00 hrs like AITool is capable of?
 
  • Like
Reactions: balucanb
Curious, is the BI integrated setup able to send a Pushover notifaction if a person is detected that is bigger than 2% of the image between 23:00:00-05:00:00 hrs like AITool is capable of?
To the best of my knowledge it does not and presently does not support many other features that AITOOL has. In fairness though BI is improving all the time albeit has a long way to go.