Version 6.0.1

Looks like a few others are seeing very high BI (w/AI turned on) RAM consumption. Mine was hovering around 9 Gigs, BUT I'm also seeing Services Host: Network Service (DNS Client) using 12 Gigs of ram. I'm not sure how BI could be causing this; however I moved back to using Blue Onyx AI and my Services Host: Network Service (DNS Client) is hovering around 350mbs and BI around 1 Gig,...

Anyone else seeing this?
 
My BI 6.0.1.4 is sitting about 5GB and is running 6 cameras with of them using the internal AI, when I first start BI the RAM is very low about 450MB, but as soon as i get an AI motion event it jumps up to around 5GB and appears to be stable at this RAM usage! This RAM usage is about 21% of my available RAM on this system (64GB installed) My system shows Services Host: Network Service (DNS Client) as using about 3 MB of memory!
 
  • Like
Reactions: ServiceXp
Am I the only one having issues with replay in web UI3, in the new version(s)? When scrolling the timeline, the app crashes after a few stop & go's and has to be relaunched manually. Log only states "Restarted after unexpected shutdown".. It doesn't happen with 5.9.9.99 :|
I'm not using ui3 but the app is doing the doing the same thing. Locks up during replay, stutters, locks up some more. Had to go back to 6.0.0.1.
 
Just took the plunge and installed 6.0.1.8, and it seems to have gone well for the most part.

The biggest issue I have is that I can't seem to get the built-in AI to work. I downloaded/installed the models at C:\BlueIris\AI\models, and then shut down CPAI and switched over to the built-in models on the general AI page. It doesn't seem to find anything upon an alert. Switching back to the CPAI models and restarting the server works, as it did before.

Since it's usually a case of user error... What am I missing?
 
Just took the plunge and installed 6.0.1.8, and it seems to have gone well for the most part.

The biggest issue I have is that I can't seem to get the built-in AI to work. I downloaded/installed the models at C:\BlueIris\AI\models, and then shut down CPAI and switched over to the built-in models on the general AI page. It doesn't seem to find anything upon an alert. Switching back to the CPAI models and restarting the server works, as it did before.

Since it's usually a case of user error... What am I missing?
I had this issue at first as well. Once I un-installed CPAI completely and folders. Then restart computer. Select "No GPU" and restart BI. If you have a GPU then select it and restart BI. It took a few times but eventually it took and has worked ever since.
 
  • Like
Reactions: Flintstone61
Speaking of the built in AI, what exactly is the purpose of the "periodic AI scan" setting? What is it doing exactly. I mean it appears to be recognizing things in the scene that are staionary like chairs, tables cars, etc. How is this helping detection overall?
 
Speaking of the built in AI, what exactly is the purpose of the "periodic AI scan" setting? What is it doing exactly. I mean it appears to be recognizing things in the scene that are staionary like chairs, tables cars, etc. How is this helping detection overall?
I created a Blue Iris help chatbot, here is the response from it


Purpose of AI Scan

what exactly is the purpose of the "periodic AI scan" setting?

Bot


The "periodic AI scan" setting in Blue Iris serves several purposes, primarily aimed at making the motion sensor "smarter" and reducing false alerts by leveraging Artificial Intelligence.

Here are the key aspects of its purpose:

  • Finding Static (Non-Moving) Objects: Originally, this scan was designed to identify static objects in a scene so that they could be ignored by alert confirmation. This helps prevent static elements from continuously triggering motion detection.
  • Becoming a Trigger Source: The periodic AI scan has evolved to the point where it can independently act as a source of triggers and alerts.
  • Timed Schedule and Dynamic Scanning:
  • By default, the scene is analyzed every 120 seconds.
  • However, if the scene changes (objects appear, move, or disappear), the scan frequency increases to every 5 seconds.
  • An object must be present or missing for a specified "make/break time" (a number of seconds) before it is considered static or removed from the static object watchlist.
  • Defining Rules: You can define rules to be applied to the static object array as a result of these scans, which can then trigger or alert the camera.
  • The trigger source for these events will be shown as "AI" in the database.
  • You can specify object labels (with wildcards allowed) and a minimum confidence for these rules.
  • The "Regard as NOT static" action can prevent specified objects from ever being considered static, overriding the normal static object functionality.
  • Rules can trigger the camera or immediately alert in various ways. Options like "trigger set" and "trigger reset" should be used carefully as they override the normal trigger break time and must be matched.
  • You can set a minimum time for an object to exist (abandoned) or not exist (missing) before an action is fired.
  • "Bidirectional" rules can detect objects as they both appear and disappear, firing once. For "exist/not exist" senses, repetitive firing of the trigger or alert can be chosen.
  • Optional Memo: An optional memo can be specified for the trigger or alert generated by a periodic AI scan.


In summary, the periodic AI scan provides a layer of intelligent analysis to constantly monitor the camera's view for object presence or absence, allowing for more sophisticated and precise triggering and alerting behaviors beyond traditional motion detection.
 
Last edited:
I created a Blue Iris help chatbot, here is the response from it


Purpose of AI Scan

what exactly is the purpose of the "periodic AI scan" setting?

Bot


The "periodic AI scan" setting in Blue Iris serves several purposes, primarily aimed at making the motion sensor "smarter" and reducing false alerts by leveraging Artificial Intelligence.

Here are the key aspects of its purpose:

  • Finding Static (Non-Moving) Objects: Originally, this scan was designed to identify static objects in a scene so that they could be ignored by alert confirmation. This helps prevent static elements from continuously triggering motion detection.
  • Becoming a Trigger Source: The periodic AI scan has evolved to the point where it can independently act as a source of triggers and alerts.
  • Timed Schedule and Dynamic Scanning:
  • By default, the scene is analyzed every 120 seconds.
  • However, if the scene changes (objects appear, move, or disappear), the scan frequency increases to every 5 seconds.
  • An object must be present or missing for a specified "make/break time" (a number of seconds) before it is considered static or removed from the static object watchlist.
  • Defining Rules: You can define rules to be applied to the static object array as a result of these scans, which can then trigger or alert the camera.
  • The trigger source for these events will be shown as "AI" in the database.
  • You can specify object labels (with wildcards allowed) and a minimum confidence for these rules.
  • The "Regard as NOT static" action can prevent specified objects from ever being considered static, overriding the normal static object functionality.
  • Rules can trigger the camera or immediately alert in various ways. Options like "trigger set" and "trigger reset" should be used carefully as they override the normal trigger break time and must be matched.
  • You can set a minimum time for an object to exist (abandoned) or not exist (missing) before an action is fired.
  • "Bidirectional" rules can detect objects as they both appear and disappear, firing once. For "exist/not exist" senses, repetitive firing of the trigger or alert can be chosen.
  • Optional Memo: An optional memo can be specified for the trigger or alert generated by a periodic AI scan.


In summary, the periodic AI scan provides a layer of intelligent analysis to constantly monitor the camera's view for object presence or absence, allowing for more sophisticated and precise triggering and alerting behaviors beyond traditional motion detection.
Nearly all of that bot's explanation is directly available in the BI help file. Regardless, in actual practice, I haven't found this feature particularly useful: in one of my camera views, at this time of the year only, my holiday icicle lights very slightly obscure the views across my lawn and driveway. The parked cars get correctly identified by this "static object" feature, but all it takes is the light wind movement of the hanging icicle lights to trip up this identification and send motion alerts for the vehicles, even though they haven't moved at all. So much for the A.I.....
 
  • Like
Reactions: Knocker
Nearly all of that bot's explanation is directly available in the BI help file. Regardless, in actual practice, I haven't found this feature particularly useful: in one of my camera views, at this time of the year only, my holiday icicle lights very slightly obscure the views across my lawn and driveway. The parked cars get correctly identified by this "static object" feature, but all it takes is the light wind movement of the hanging icicle lights to trip up this identification and send motion alerts for the vehicles, even though they haven't moved at all. So much for the A.I.....
I believe part of the beauty of this feature is adding things like your icicle lights to be ignored in a rule. Much like tree branches and other stationary objects than can create motion at times from wind etc.
 
I believe part of the beauty of this feature is adding things like your icicle lights to be ignored in a rule. Much like tree branches and other stationary objects than can create motion at times from wind etc.
Sounds good in theory, but in practice the hanging icicle light strings aren't "seen" or identified at all, until they are blowing around, in which case CPAI (or the new built-in BI AI) will sometimes identify them as a person, which is really unhelpful, or, as I said before, will trick the system into imagining motion of the (static, motionless) parked vehicles.
 
I just switched on all of my cams to Built in AI for tinkering sake. Some variability in resource usage as usual.
1767170990266.png
1767171016474.png
 
  • Like
Reactions: gwminor48
Sounds good in theory, but in practice the hanging icicle light strings aren't "seen" or identified at all, until they are blowing around, in which case CPAI (or the new built-in BI AI) will sometimes identify them as a person, which is really unhelpful, or, as I said before, will trick the system into imagining motion of the (static, motionless) parked vehicles.
I do love my Christmas lights, but always glad when they come down and my camera's go back to normal, they are quite a challenge. In my case, I just find it easier to mask them out completely.
 
I just find it easier to mask them out completely.
I have permanent holiday lights on my house that come on for various holidays and occasions. Like you, I found it challenging to get my cameras to play well with my lights. I ended up creating a "holiday" profile in BI which basically disables the alerts on the cameras in the affected areas. To take this one step further, I have my home automation system switch my BI profiles based on the status of my light controller. If the lights are on, the cameras in the affected areas stop recording alerts but still record a continuous stream just in case I need to review something later.
 
I have permanent holiday lights on my house that come on for various holidays and occasions. Like you, I found it challenging to get my cameras to play well with my lights. I ended up creating a "holiday" profile in BI which basically disables the alerts on the cameras in the affected areas. To take this one step further, I have my home automation system switch my BI profiles based on the status of my light controller. If the lights are on, the cameras in the affected areas stop recording alerts but still record a continuous stream just in case I need to review something later.
I didn't realize I had started an entire sub-topic of discussion in regards to holiday lights and getting our cameras to deal with those, LOL. Yes, I will also be moving to permanent holiday lights as well next year (purchased, but didn't have enough time yet for this season, open soffit complications). I have already been running a "winter/holiday" schedule in BI for a few years: it significantly cuts down on the BI motion detection sensitivity and regions of concern, and relies much more heavily on the in-camera ONVIF IVS capabilities. The AI of the cameras (Dahau S2 5442s) does a pretty amazing job of filtering out most (but definitely not all) of the false triggers, but in this "holiday" setup I do lose pretty much all hope of any "wildlife" captures with the reduced BI (CPAI or newer built-in AI) sensitivity.
 
OK... Folders were renamed/moved, system rebooted, and still no dice. Gone between CPU only and GPU, lots of saves, reboots, restarts, but still doesn't seem to find anything when triggered.

Anyone have any advice?
Can you post your settings of the camera and the BI/settings/AI?

Screenshot 2025-12-31 123910.jpg Screenshot 2025-12-31 124100.jpg