AI motion detection with BlueIris: an alternate approach

As many others here I am tired of getting hundreds of false alarms from trees moving in the wind and insects. I am about to try this solution that seems to do exactly what everybody wants. As a start, I have upgraded Win7 to Win10 and BI4 to BI5. Before I go ahead, I have a question: If I log into BI remotely with UI3 just to check things out when there are no triggers, can I then see the full resolution stream or will I have to accept the low 640 x 480 resolution?
 
As many others here I am tired of getting hundreds of false alarms from trees moving in the wind and insects. I am about to try this solution that seems to do exactly what everybody wants. As a start, I have upgraded Win7 to Win10 and BI4 to BI5. Before I go ahead, I have a question: If I log into BI remotely with UI3 just to check things out when there are no triggers, can I then see the full resolution stream or will I have to accept the low 640 x 480 resolution?

Technically it's up to you how you do it. You could record full HD all the time and send HD images to DeepStack, but that's a bit slow (3-6 seconds on my i9 per image). I started out this way and said screw it and did a constantly recording SD stream which also sends pics to DeepStack and then set a generous pre-buffer time for HD stream activity based recording.

For some reason 5 to 10 seconds often missed the motion portion.

I recently after a month of use upped my threshold levels based on what I've been seeing, which dramatically reduces false alerts.
 
  • Like
Reactions: t_andersen
As many others here I am tired of getting hundreds of false alarms from trees moving in the wind and insects. I am about to try this solution that seems to do exactly what everybody wants. As a start, I have upgraded Win7 to Win10 and BI4 to BI5. Before I go ahead, I have a question: If I log into BI remotely with UI3 just to check things out when there are no triggers, can I then see the full resolution stream or will I have to accept the low 640 x 480 resolution?

If you set the cameras up as described on the wiki you'll get full HD streams when you hit UI3 to view both recorded and live footage.
 
  • Like
Reactions: t_andersen
You could record full HD all the time and send HD images to DeepStack, but that's a bit slow (3-6 seconds on my i9 per image).
That's very slow. I use the Windows version on my i5-8500T 2.1Ghz and I send 2048x1536 resolution files to Deepstack which is analysed in 5-800ms.

I run Blue iris with 4 HikVision cameras on the same machine.
 
Version 5.2.1 is now available. This release addresses two minor security vulnerabilities in dependent libraries. There are no functional changes.

To update run docker-compose pull then docker-compose up to restart.
 
  • Like
Reactions: T_K
Hoping someone can help with my problem. I'm using DeepquestAI running on the same computer that BI runs, Windows 10, i7 6700 with 48 gigs RAM, and the AI Tool version 1.67 but I wanted to offload this computer which hovers between 85 and >95% CPU utilization most of the time (I have 7 HD cameras and 8 UHD cameras).

I installed Docker on a separate computer running Ubuntu Desktop 20.04 LTS, 4 gig RAM, Intel Celeron CPU. Docker install went well, no errors, then installed Deepstack which also went well and runs seemingly ok.

However, when I point AI Tool to this Linux computer I'm getting these errors

UFW is not enabled on the Lunix box.

Any help or suggestions is appreciated.




COMMAND TO START DEEPQUEST AI ON UBUNTU
sudo docker run -e VISION-SCENE=True -v localstorage:/datastore -p 82:5000 deepquestai/deepstack



DEEPQUEST AI SCREEN OUTPUT
/v1/vision/scene
---------------------------------------
v1/vision/addmodel
---------------------------------------
v1/vision/listmodels
---------------------------------------
v1/vision/deletemodel
---------------------------------------
---------------------------------------
v1/backup
---------------------------------------
v1/restore
[GIN] 2020/08/02 - 17:53:15 | 403 | 274.22µs | 192.168.1.24 | POST /v1/vision/detection
[GIN] 2020/08/02 - 17:53:15 | 403 | 72.503µs | 192.168.1.24 | POST /v1/vision/detection
[GIN] 2020/08/02 - 17:53:15 | 403 | 59.593µs | 192.168.1.24 | POST /v1/vision/detection
[GIN] 2020/08/02 - 17:53:15 | 403 | 62.307µs | 192.168.1.24 | POST /v1/vision/detection


[02.08.2020, 11:56:35.953]:
[02.08.2020, 11:56:35.965]: Starting analysis of G:\aiinput\GTHSD.20200802_115616406.jpg
[02.08.2020, 11:56:35.975]: (1/6) Uploading image to DeepQuestAI Server
[02.08.2020, 11:56:37.997]: System.Net.Http.HttpRequestException | An error occurred while sending the request. (code: -2146233079 )
[02.08.2020, 11:56:38.007]: ERROR: Processing the following image 'G:\aiinput\GTHSD.20200802_115616406.jpg' failed. Can't reach DeepQuestAI Server at .


[02.08.2020, 11:56:38.019]:
[02.08.2020, 11:56:38.028]: Starting analysis of G:\aiinput\BGDSD.20200802_115616486.jpg
[02.08.2020, 11:56:38.040]: (1/6) Uploading image to DeepQuestAI Server
[02.08.2020, 11:56:40.055]: System.Net.Http.HttpRequestException | An error occurred while sending the request. (code: -2146233079 )
[02.08.2020, 11:56:40.067]: ERROR: Processing the following image 'G:\aiinput\BGDSD.20200802_115616486.jpg' failed. Can't reach DeepQuestAI Server at .
 
Can anyone who's gotten this to work send a copy of their triggers.json file? I'm having difficulting understanding how to modify mine using the example.

Thanks.
 
There is mine... I hope this help.

1596482953672.png
 
Is there a way to see the object that met the trigger criteria like in the history tab in AI Tools? This morning I'm suddenly getting motion emails from BI but upon review I don't see any motion and I'm not sure what is triggering BI.
 
Is there a way to see the object that met the trigger criteria like in the history tab in AI Tools? This morning I'm suddenly getting motion emails from BI but upon review I don't see any motion and I'm not sure what is triggering BI.

The system logs everything. Run Docker Desktop and look at the log output from the trigger container. It'll show what was detected, the confidence level etc.
 
My setup using neile's all docker method is taking approximately 35 seconds per file. That sounds way too long to me and as BI is depositing an image every 5 seconds the simple math means it will be getting further and further behind especially with 3 cameras being analyzed. Am I missing or misreading something in the logs?

i7 6700 with 48 gig RAM. CPU utilization hovers around 70%
 
My setup using neile's all docker method is taking approximately 35 seconds per file. That sounds way too long to me and as BI is depositing an image every 5 seconds the simple math means it will be getting further and further behind especially with 3 cameras being analyzed. Am I missing or misreading something in the logs?

i7 6700 with 48 gig RAM. CPU utilization hovers around 70%

Are you running the cameras with direct-to-disk recording? If not you should be.

Also see the troubleshooting wiki for how to resolve high CPU usage.
 
Yes, I've been doing direct-to-disk for a few years now. With 15 cameras it's pretty mandatory.
I don't see any of the issues outlined in the link you provided. Here is an excerpt of my logs:


2020-08-04T09:29:36-06:00 [Trigger Garage to House Camera] /aiinput/GTHSD.20200804_092935648.jpg: Analyzing
2020-08-04T09:29:37-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092936002.jpg: Analyzing
2020-08-04T09:29:37-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092936282.jpg: Analyzing
2020-08-04T09:29:37-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092125905.jpg: No objects detected. (491191 ms)
2020-08-04T09:29:40-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092126281.jpg: No objects detected. (493561 ms)
2020-08-04T09:29:40-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092940282.jpg: Analyzing
2020-08-04T09:29:43-06:00 [Trigger Garage to House Camera] /aiinput/GTHSD.20200804_092941636.jpg: Analyzing
2020-08-04T09:29:43-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092942059.jpg: Analyzing
2020-08-04T09:29:44-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092130285.jpg: No objects detected. (493407 ms)
2020-08-04T09:29:46-06:00 [Trigger Garage to House Camera] /aiinput/GTHSD.20200804_092945581.jpg: Analyzing
2020-08-04T09:29:47-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092946053.jpg: Analyzing
2020-08-04T09:29:47-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092946294.jpg: Analyzing
2020-08-04T09:29:48-06:00 [Trigger Garage to House Camera] /aiinput/GTHSD.20200804_092131958.jpg: No objects detected. (496460 ms)
2020-08-04T09:29:50-06:00 [Trigger Driveway North Camera] /aiinput/DWNSD.20200804_092950297.jpg: Analyzing
2020-08-04T09:29:51-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092131932.jpg: No objects detected. (499768 ms)
2020-08-04T09:29:53-06:00 [Trigger Garage to House Camera] /aiinput/GTHSD.20200804_092951706.jpg: Analyzing
2020-08-04T09:29:53-06:00 [Trigger South Road Tarmac Camera] /aiinput/SRTSD.20200804_092952053.jpg: Analyzing