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

Tinbum

Pulling my weight
Joined
Sep 5, 2017
Messages
420
Reaction score
111
Location
UK
Thanks @Chris Dodge. I tried both and neither changed my display.

Also, how do I get images back in my History tab? Somehow I must have done something to prevent them from showing. Thanks
Have you tried making the window wider?
 

Senor Pibb

Young grasshopper
Joined
May 22, 2020
Messages
43
Reaction score
18
Location
Greer, SC
Also does 1.67.8.33855 respect the "camera" folder with the "older" style masks? I ask because I am not getting alerts on cars/trucks on the road. When I view history and choose show mask the "older styled" mask do appear. But still getting alerted.
 

105437

BIT Beta Team
Joined
Jun 8, 2015
Messages
1,026
Reaction score
224
@105437 - hmm. So are you running AITOOLS through a remote viewer of any sort? If so can you test directly on the screen? What is your windows display DPI set to? right-click desktop > display settings > scaling. How is the monitor connected? If your monitor has been disconnected or had a resolution change during the current windows session, try a reboot.

The missing images is because the vertical splitter bar that allows you to resize between the list and the image is missing/offscreen.

All window and splitter positions are remembered in the registry. Close the app, then go to Start menu > Run > Regedit > browse to this and delete: HKEY_CURRENT_USER\SOFTWARE\AITool
You are correct. I use Remote Desktop Connection to remotely view the Windows 10 PC where I run everything. I changed the video setting for this connection to 1080p and it allowed me to see the vertical resize bar on the History tab and the UI oddities were corrected. Thanks!
 

Chris Dodge

Pulling my weight
Joined
Aug 9, 2019
Messages
77
Reaction score
102
Location
massachusetts
Also does 1.67.8.33855 respect the "camera" folder with the "older" style masks? I ask because I am not getting alerts on cars/trucks on the road. When I view history and choose show mask the "older styled" mask do appear. But still getting alerted.
It should still be respecting the older masks in \camera\camername.bmp/png. Can you elaborate a bit? Does a 'relevant' alert appear in the history list for something that should be masked out?
 

Senor Pibb

Young grasshopper
Joined
May 22, 2020
Messages
43
Reaction score
18
Location
Greer, SC
It should still be respecting the older masks in \camera\camername.bmp/png. Can you elaborate a bit? Does a 'relevant' alert appear in the history list for something that should be masked out?
Yes here is one example:
Added a another example.

The rock looks like a person to the AI. I added a mask over it. But it is still red, still learning the new interface. But I think that red means active not masked?
 

Attachments

Last edited:

juched

n3wb
Joined
Sep 10, 2020
Messages
15
Reaction score
5
Location
Waterloo, ON
@105437 @juched @Senor Pibb - For the odd display issues, make a shortcut to AITOOLS somewhere, right-click > properties > compatibility tab > change high dpi settings, and see if they help for the next time you start it. Or maybe edit aitools.exe.config and set the value for DpiAwareness to "false" rather than "PerMonitorV2". I have a 4k monitor set at 175% dpi and a 1080p monitor set at 100% and UI elements seem to be fine for both.
Thank you. I have tried all the various options already under High DPI settings, and none of them work. If I open the window While the TV is on, it works, but if I close the app and relaunch once the the TV is off, then it will not draw properly. Yes, when I turn the TV off the resolution changes, even if my TeamViewer is open while the TV turns off. So, it seems the High DPI setting doesn't work when a monitor is not attached (or is off).

Have not had the opportunity to try DpiAwareness option yet.
 

pmcross

Pulling my weight
Joined
Jan 16, 2017
Messages
370
Reaction score
184
Location
Pennsylvania
@Chris Dodge I am having an issue where AI Tool isn't monitoring the input folder. I am running the latest version. I have it defined under the settings tab. Do I also need to set each input folder on each camera in AI Tool?
 

Chris Dodge

Pulling my weight
Joined
Aug 9, 2019
Messages
77
Reaction score
102
Location
massachusetts
@Chris Dodge I am having an issue where AI Tool isn't monitoring the input folder. I am running the latest version. I have it defined under the settings tab. Do I also need to set each input folder on each camera in AI Tool?
You dont need a path per camera, but make sure any paths are either blank or valid paths. Turn on LOG EVERYTHING, hit SAVE on settings tab, hop right over to the log tab and let me know any recent log entries from "UpdateWatchers". It should give you an error if it cant enable a path for some reason. Also, over on the Camers tab, verify all the objects are still checked. Have seen a few cases they get unchecked.
 
Joined
Sep 21, 2020
Messages
5
Reaction score
2
Location
UK
I'm also having the display issues with the latest version (181). For example in the history tab, whatever I do with the window size and dividers I can never see the whole history image. I'm connecting via remote VNC. I can temporarily change the resolution to 1080 (which upsets the local monitor), then if I resize the AITool window I can change back to my normal remote resolution and the windows and dividers can be adjusted to fit the whole image in. The latest changes to the history tab now I can see them correctly are great. So a temporary fix is to change to a high resolution, move the window edges and dividers then change the resolution back again. Hope this helps.
 

Alan_F

n3wb
Joined
May 17, 2019
Messages
14
Reaction score
6
Location
Maryland
1. In the next release in a day or two, close the app, edit aitool.settings.json in notepad and search for "action_mqtt_retain_message" and set to true for each camera you need.
2. History tab has been overhauled for the next release too. It should fix missing items, and will actually show you skipped images due to cooldown or other failure.
I learned the hard way that if you a flagging an alert using MQTT and set the retain to true it will flag every following alert also. I had set the retain to True in the config file back before it had any effect, and with the new version I noticed that each time the AI detected a person all of the alerts were flagged with that detection until the next detection occurred, and then everything following was flagged with that detection, and so on. I'm sure there are some uses for the retain flag, but you probably don't want to use it if one of your MQTT topics is telling Blue Iris to flag alerts.
 
Joined
Oct 1, 2020
Messages
16
Reaction score
2
Location
Texas
So I have been having quite a few issues lately with DeepStack running in an Ubuntu VM... At first it was running pretty smoothly, less than 1 second in the queue even with just one instance running. Well, I am now at over 9000ms even with two instances running! The VM has access to 24 cores on two Xeon X5650 CPUs, no AVX sadly. It is also eating RAM at about 13GBs right now. Reading that a few have managed under 1000ms times on a Pi4 and NCS2 I was thinking I might give that a shot.. until I just thought, any reason I can't just use something like the NCS2 on the PC? At that point, would there be any better options? The host is a Dell R710 so not many options for a GPU, unless it didn't need external power.
 

pmcross

Pulling my weight
Joined
Jan 16, 2017
Messages
370
Reaction score
184
Location
Pennsylvania
You dont need a path per camera, but make sure any paths are either blank or valid paths. Turn on LOG EVERYTHING, hit SAVE on settings tab, hop right over to the log tab and let me know any recent log entries from "UpdateWatchers". It should give you an error if it cant enable a path for some reason. Also, over on the Camers tab, verify all the objects are still checked. Have seen a few cases they get unchecked.
This is working now. I believe that the problem was that I had a AI Tool service pointing to the exe and also had to console window open. This used to work on GentlePumpkin's tool. I am really enjoying the ability to draw a mask for each camera directly from your tool as well as the dynamic masking. Thank you very much for your hard work and development of the this fork.
 

pmcross

Pulling my weight
Joined
Jan 16, 2017
Messages
370
Reaction score
184
Location
Pennsylvania
So I have been having quite a few issues lately with DeepStack running in an Ubuntu VM... At first it was running pretty smoothly, less than 1 second in the queue even with just one instance running. Well, I am now at over 9000ms even with two instances running! The VM has access to 24 cores on two Xeon X5650 CPUs, no AVX sadly. It is also eating RAM at about 13GBs right now. Reading that a few have managed under 1000ms times on a Pi4 and NCS2 I was thinking I might give that a shot.. until I just thought, any reason I can't just use something like the NCS2 on the PC? At that point, would there be any better options? The host is a Dell R710 so not many options for a GPU, unless it didn't need external power.
I am running a Raspberry Pi 4b with an NCS 2 and it is working very well for me. I also was interested in running the NCS on a PC, but the problem is that Deepstack for the NCS is only supported on a Raspberry Pi.
 

Tinbum

Pulling my weight
Joined
Sep 5, 2017
Messages
420
Reaction score
111
Location
UK
I am running a Raspberry Pi 4b with an NCS 2 and it is working very well for me. I also was interested in running the NCS on a PC, but the problem is that Deepstack for the NCS is only supported on a Raspberry Pi.
I have 5 pi and I cant get any of them to show the box round the object in the correct place. I cant understand why. Each has a brand new sd card with a new individual intall. Ive tried full, lite and the desktop lite versions and the 64bit. None show the madk in the correct place. Any ideas?
 

pmcross

Pulling my weight
Joined
Jan 16, 2017
Messages
370
Reaction score
184
Location
Pennsylvania
I have 5 pi and I cant get any of them to show the box round the object in the correct place. I cant understand why. Each has a brand new sd card with a new individual intall. Ive tried full, lite and the desktop lite versions and the 64bit. None show the madk in the correct place. Any ideas?
I'm honestly not sure. My boxes are around objects properly. I used this to get Deepstack installed on my pi:

The only difference was that I used this command to install
sh install-deepstack.sh
 

Chris Dodge

Pulling my weight
Joined
Aug 9, 2019
Messages
77
Reaction score
102
Location
massachusetts
Thats what I've done numerous times from scratch. It happens with all versions of AITool.
@Tinbum @pmcross - Interesting! The same version of deepstack, one works fine. Sooo... How do you both install the OS? Are there not different versions available outside lite/full/desktop? Did you both update the OS to the latest version after install? How are you connecting to the device to manage it? Is there an actual monitor attached, is it on, what kind of remote tool are you using, etc - The deepstack code could be using some dpi/resolution setting set on the device based on what is connected or not. That would be bad, but still could happen. Unrelated, Is anyone else a little pissed that ipcamtalk is limiting me to ONE SPACE after a period? Dont tell me how to live my life, ipcamtalk. :)
 

pmcross

Pulling my weight
Joined
Jan 16, 2017
Messages
370
Reaction score
184
Location
Pennsylvania
@Tinbum @pmcross - Interesting! The same version of deepstack, one works fine. Sooo... How do you both install the OS? Are there not different versions available outside lite/full/desktop? Did you both update the OS to the latest version after install? How are you connecting to the device to manage it? Is there an actual monitor attached, is it on, what kind of remote tool are you using, etc - The deepstack code could be using some dpi/resolution setting set on the device based on what is connected or not. That would be bad, but still could happen. Unrelated, Is anyone else a little pissed that ipcamtalk is limiting me to ONE SPACE after a period? Dont tell me how to live my life, ipcamtalk. :)
This is a great point @Chris Dodge I used the Raspberry Pi Imager and went with the full version. I then updated the OS and installed Deepstack. I connect via VNC Viewer and don't have a monitor connected to the Pi.

Edit: it may be important to note that I use RDP to connect to my BI machine to view/configure AI Tool.
 
Top