[tool] [tutorial] Free AI Person Detection for Blue Iris
- Thread starter GentlePumpkin
- Start date
Don_TranQuil
n3wb
For some reason that error is irrelevant, it still works normal.
No issue's yet. so i think thats only if u using it on the same pc only that it gathers the output folders.
No issue's yet. so i think thats only if u using it on the same pc only that it gathers the output folders.
Don_TranQuil
n3wb
Is it on the same machine?
well i don't use that, as thats only to detect the folders.
i have a same error while not using it on my same server.. i ignore it and it still works.
what this thing does is already populate the folders to capture in the aitool.
and ofcourse use that ip for the variables
u get only that error if u press the save button at the page where u put the ip of blue iris
just manual add the folders to monitor and it works normal
New version of AiTools out on GitHub.
A few minor changes:
A few minor changes:
- Feature: In Relevant objects, you can now independently control if you ignore Image or Dynamic masks.
- Anything other than the main/default Relevant object list will have its objects min/max threshold and ignore mask settings restricted by the same object found in the main camera list (the gatekeeper). Also Ignore image mask and dynamic mask will be grayed out if it is not the main cameras mask
bluehawk
n3wb
Aitools - Node-red - Chromecast
I have a permanent monitor setup that displays the camera feed of an alert-trigged camera. This works when Node-red sends a request to the monitor's Chromecast if it receives an MQTT alert message from AITools. It also sends a TTS message to various Google speakers announcing the event.
This works great except when someone or something hangs around on camera for too long and causes multiple alerts in a short time frame. For instance, my driveway camera will trigger multiple times for the same car or my dog will cause multiple alerts when she is on the lawn camera. This can get irritating with multiple announcements on the speakers and the monitor video constantly getting replaced by a new feed.
I am wondering if it is best to attack this problem programmatically in Node-red or if there if are settings in BlueIris that can deal with this.
I would also, as a first-time poster, like to thank the authors of AITools for such a great job and the many posters here who have shared so much helpful info. My apologies if this topic has already been covered. I did/am attempting to read the entire thread but it is just way too long.
Cheers
I have a permanent monitor setup that displays the camera feed of an alert-trigged camera. This works when Node-red sends a request to the monitor's Chromecast if it receives an MQTT alert message from AITools. It also sends a TTS message to various Google speakers announcing the event.
This works great except when someone or something hangs around on camera for too long and causes multiple alerts in a short time frame. For instance, my driveway camera will trigger multiple times for the same car or my dog will cause multiple alerts when she is on the lawn camera. This can get irritating with multiple announcements on the speakers and the monitor video constantly getting replaced by a new feed.
I am wondering if it is best to attack this problem programmatically in Node-red or if there if are settings in BlueIris that can deal with this.
I would also, as a first-time poster, like to thank the authors of AITools for such a great job and the many posters here who have shared so much helpful info. My apologies if this topic has already been covered. I did/am attempting to read the entire thread but it is just way too long.
Cheers
You can adjust your cooldown period for triggering BI and for sending alerts within AiTool. For the dog you can schedule when you want to trigger for dogs.
I send my triggers to HomeSeer via MQTT but I thought about sending to Node-Red as my first Node-Red project. What I do that my be helpful for you is I have devices created in HomeSeer for each object I am interested in (vehicles, animals, person). I group the objects by camera. I then create events to limit the false notifications such as ignore the dogs during the daytime only alerting when an animal is detected at night or only send out an alert every 15 minutes when a vehicle is detected. I have different criteria for when to trigger BI to record and when to send out a text or push notification.
Can someone help me iron out this issue i'm having? I get these 0s flags pretty often. I'm sure it's setting related but I cant figure out which combination of settings doesn't line up correctly.
I suspect it is finding a different amount of relevant objects in that time period and re-triggers so maybe I dont understand dynamic masking?
If I look at the same clip in UI3 web version it's a 10s clip that overlaps the previous one:
Relevant info:
In BI i'm doing continuous recording, 1 camera method.
Settings In BI:
Trigger tab: 8.0s break time
JPEG every 1.0s
7.0s pre-trigger
Settings in AI Tool:
dynamic masking w/ detail settings checked
0.3minute cooldown
Triggers URLs:
Cancel URL:
I suspect it is finding a different amount of relevant objects in that time period and re-triggers so maybe I dont understand dynamic masking?
If I look at the same clip in UI3 web version it's a 10s clip that overlaps the previous one:
Relevant info:
In BI i'm doing continuous recording, 1 camera method.
Settings In BI:
Trigger tab: 8.0s break time
JPEG every 1.0s
7.0s pre-trigger
Settings in AI Tool:
dynamic masking w/ detail settings checked
0.3minute cooldown
Triggers URLs:
Cancel URL:
Last edited:
bluehawk
n3wb
Thank you for this I did not see that setting before.
My "cooldown"s are all set to zero, so hopefully that will fix it.
Should I use any of the BI settings as well? ie under:
Alert/Timer
- Minimum time between alerts -
- Min/max time between trigers?
Mine are all turned off.
Your HomeSeer ideas are interesting have to think about doing something like that in Node-Red.
I use Hubitat for my H.A. backend. I did not like the Hubitat "Rule Machine" app so I switched to using Node-red recently for more complex routines. I now love Node-red. It opens up a whole new wold of possibilities for me. It is also fun to use. Not sure how useful it would be to HomeSeer users though.
I have HomeSeer and use Node-Red almost exclusively to control my HomeSeer devices. I do it via MQTT (If you're not familiar with it, read up on it. Not too hard to install a "server" for it. It's a simple message queue system). There is a free plugin for HomeSeer called mcsMQTT which allows you to associate devices in HomeSeer with MQTT messaging. I control everything using MQTT via Node-Red. It works very well. If anyone would like more details, just let me know.
Last edited:
bluehawk
n3wb
I don't use those settings in BI but I believe you could use the first setting to control how long to wait after an alert after an alert is sent (AiTool Cooldown should do same). The min/max time between triggers controls alert between multiple cameras.
I just noticed this today in Blue Iris and It seems to only support Windows deepstack. The help file does not mention AI Tool. The options seem limited. .
I started experimenting with it last night and while it is nice to have direct integration, its still limited compared to AI Tools. Its the first release, so i'm hopeful we'll see more features/refinements over time, but its a great start. Finally something that doesn't add another subscription, yet brings value.
Keep in mind that AI Tools and many of these other AI platforms are based off of the Deepstack engine.
Right now BI only brings us Deepstack implementation. In time it may bring over AI Tools or another one, or BI may create their own set of tools to run native to BI, or it may say this is sufficient for the majority of the users...
Right now BI only brings us Deepstack implementation. In time it may bring over AI Tools or another one, or BI may create their own set of tools to run native to BI, or it may say this is sufficient for the majority of the users...
Eastwood
n3wb
- Mar 14, 2018
- 21
- 0
Working on reading this thread.. what are the big things AI Tools does that BI's implementation doesn't? Is it possible to have it ignore my cars??
AiTools offers these main features.
Options for the program to dynamically mask out cars once parked until moved.
You can create a static mask for an area to ignore certain objects. Both types of masks have scheduling capabilities.
You can set detection thresholds for each object type.
MQTT topic and payload customization.
Thresholds can be different for for each type of alert. Detection thresholds can also be schedule based.
Can use other AI analysis services other than DeepStack or in combination with DeepStack.
Hello!
Very happy to have this resource. I've been working on it for a few days and can't seem to get past this point. Seems Python is giving an error at startup. I can't find that file it's looking for either. Did I miss a setup step for Python?
Thank you
Debug: DeepStack>> Init: File "../init.py", line 9, in <module>
DeepStack>> C:\DeepStack\interpreter\python.exe: can't open file '../intelligence.py': [Errno 2] No such file or directory
Very happy to have this resource. I've been working on it for a few days and can't seem to get past this point. Seems Python is giving an error at startup. I can't find that file it's looking for either. Did I miss a setup step for Python?
Thank you
Debug: DeepStack>> Init: File "../init.py", line 9, in <module>
DeepStack>> C:\DeepStack\interpreter\python.exe: can't open file '../intelligence.py': [Errno 2] No such file or directory