Menu
What's new

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

vonarstonehelm

vonarstonehelm

n3wb
Joined
Dec 18, 2020
Messages
3
Reaction score
0
Location
VA
Have a quick question... I have had AI tool running for a while now and would like to mask off a portion of the street and the houses across the street. I took a jpg, created a mask in gimp and saved it as a png with the same Cam1-xxxxxxx.png. Do I need to name it something special like Cam1-mask.png? I have restarted AI tool a couple times but it when I turn on "See Mask" it doesn't seem to load anything.

Can someone point me in the right direction please?

Thank you
 
M

Mattias Fornander

n3wb
Joined
Sep 17, 2019
Messages
7
Reaction score
3
Location
Boulder, CO
vonarstonehelm said:
Have a quick question... I have had AI tool running for a while now and would like to mask off a portion of the street and the houses across the street. I took a jpg, created a mask in gimp and saved it as a png with the same Cam1-xxxxxxx.png. Do I need to name it something special like Cam1-mask.png? I have restarted AI tool a couple times but it when I turn on "See Mask" it doesn't seem to load anything.

Can someone point me in the right direction please?

Thank you
Click to expand...
Do a manual trigger in BI (right click on cam, trigger) to feed at least one image into AI-Tool. Then you can apply a mask. Note that masks in AI-Tool are the opposite (ignore) to what they are in BI (include).
 
vonarstonehelm

vonarstonehelm

n3wb
Joined
Dec 18, 2020
Messages
3
Reaction score
0
Location
VA
Mattias Fornander said:
Do a manual trigger in BI (right click on cam, trigger) to feed at least one image into AI-Tool. Then you can apply a mask. Note that masks in AI-Tool are the opposite (ignore) to what they are in BI (include).
Click to expand...
Okay, so I manually trigger BI. Use that image to make the mask? I guess I am missing how I apply the mask (besides inside a photo editing software).
 
cscoppa

cscoppa

Getting the hang of it
Joined
Dec 14, 2019
Messages
50
Reaction score
26
I do my masking inside BI, I find it easier and more powerful with multiple zones. Why even send the snapshot to the AI if the motion is in an area you don't want to monitor.
 
M

Mattias Fornander

n3wb
Joined
Sep 17, 2019
Messages
7
Reaction score
3
Location
Boulder, CO
cscoppa said:
I do my masking inside BI, I find it easier and more powerful with multiple zones. Why even send the snapshot to the AI if the motion is in an area you don't want to monitor.
Click to expand...
I do that too but if I send the whole image to AI-Tool it picks up all kinds of false objects and they are usually the same objects.. My pots are mistaken for people in certain light conditions quite often. A second mask in AI-Tool prevents these false detentions.
 
cscoppa

cscoppa

Getting the hang of it
Joined
Dec 14, 2019
Messages
50
Reaction score
26
Mattias Fornander said:
I do that too but if I send the whole image to AI-Tool it picks up all kinds of false objects and they are usually the same objects.. My pots are mistaken for people in certain light conditions quite often. A second mask in AI-Tool prevents these false detentions.
Click to expand...
What I'm saying is I have the mask setup such that it DOESN'T send the image to the AITOOLS folder. I'm avoiding things like pots, etc. using the mask in BI. The image only goes into the AITOOLS folder if the motion is in an area that I want to track.
 
balucanb

balucanb

Getting the hang of it
Joined
Sep 19, 2020
Messages
147
Reaction score
23
Location
TX
Anybody with Docker/portainer/postman exp. willing to help me out with this stinking attempt at deploying the custom model. IMA bout to quit!
 
M

maximosm

Young grasshopper
Joined
Jan 8, 2015
Messages
95
Reaction score
6
Mattias Fornander said:
Are you reducing the size of your saved images? Make BI resize them to 640x480 or even 320x240 if you you're running on default Medium MODE.
[/QUO
Click to expand...
cjowers said:
-try deepstackai on 'low' mode
-try lower frequency of images being anaylized
-try lower resolution images from a substream camera stream
-try less cameras triggering deepstack
-try latest deepstack versions
-try a jetson nano

gpu will definately help, its way more efficient at this type of computation (but you have to run the gpu deppstack version)
Click to expand...
I will try all off them
-try deepstackai on 'low' mode - not realy diference
-try lower frequency of images being anaylized - this helped but not as expected
-try lower resolution images from a substream camera stream - this gives an error
-try less cameras triggering deepstack - i trigger only 2 cameras (at same time)
-try latest deepstack versions - i use the last version
----- jetson nano is very interesting. ** very interesting- is it stable?How many cameras can support version with 2GB?Can we run ubundu at this device?
 
Village Guy

Village Guy

Pulling my weight
Joined
May 6, 2020
Messages
291
Reaction score
161
Location
UK
cscoppa said:
What I'm saying is I have the mask setup such that it DOESN'T send the image to the AITOOLS folder. I'm avoiding things like pots, etc. using the mask in BI. The image only goes into the AITOOLS folder if the motion is in an area that I want to track.
Click to expand...
If I understand your suggestion you are proposing masking your pots from generating a trigger. Do your pots go walkabout?

Seriously though IMHO your methodology is flawed unless I have missed something. Masking in BI will only stop a trigger, it will not stop your pot from being recognised as a person. If the camera triggers on something else, the pot will still be in the image being analysed, this where aitool masking comes into play.
 
cscoppa

cscoppa

Getting the hang of it
Joined
Dec 14, 2019
Messages
50
Reaction score
26
Village Guy said:
If I understand your suggestion you are proposing masking your pots from generating a trigger. Do your pots go walkabout?

Seriously though IMHO your methodology is flawed unless I have missed something. Masking in BI will only stop a trigger, it will not stop your pot from being recognised as a person. If the camera triggers on something else, the pot will still be in the image being analysed, this where aitool masking comes into play.
Click to expand...
Ah, I see what you're saying. I guess in certain situations that might be useful. In my setup I'm only actually recording video clips when a human is detected. I haven't run into a situation (considering my BI masks) where I would need to do this. It doesn't bother me if the AI says "this is a pot" because it's ALSO accurately saying this is a human.
 
Village Guy

Village Guy

Pulling my weight
Joined
May 6, 2020
Messages
291
Reaction score
161
Location
UK
cscoppa said:
Ah, I see what you're saying. I guess in certain situations that might be useful. In my setup I'm only actually recording video clips when a human is detected. I haven't run into a situation (considering my BI masks) where I would need to do this. It doesn't bother me if the AI says "this is a pot" because it's ALSO accurately saying this is a human.
Click to expand...
I think you are still missing the point. Aitool is known for misinterpreting a pot or ornament for a person occasionally, especially in low light conditions. If a moth or spider or even rain triggers the camera you could get a false positive. Your pot is suddenly reported as a human.
Needless to say this is only an issue for folk that experience a lot of false positives.
 
M

maximosm

Young grasshopper
Joined
Jan 8, 2015
Messages
95
Reaction score
6
I have made a docker to a linux nuc 7 and is running faster 300mS(network server) as the new i7 10700 windows isntalation 750mS(localhost server)..... Is there anyone that uses JETSON NANO 2GB to say in how mS working?
 
T

Tinbum

Pulling my weight
Joined
Sep 5, 2017
Messages
448
Reaction score
126
Location
UK
maximosm said:
Is there anyone that uses JETSON NANO 2GB to say in how mS working?
Click to expand...
4GB


[GIN] 2020/12/19 - 14:54:48 | 200 | 622.795749ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:54:56 | 200 | 467.233481ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:01 | 200 | 603.681984ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:04 | 200 | 619.815421ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:12 | 200 | 982.65425ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:15 | 200 | 977.080009ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:18 | 200 | 614.320557ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:52 | 200 | 615.467718ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:55 | 200 | 687.367886ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:58 | 200 | 664.312052ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:11 | 200 | 690.881715ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:14 | 200 | 623.645039ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:18 | 200 | 637.22545ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:21 | 200 | 619.848391ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:24 | 200 | 682.57585ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:27 | 200 | 701.910141ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:31 | 200 | 897.778893ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:33 | 200 | 599.392303ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:36 | 200 | 688.946672ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:49 | 200 | 662.386643ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:52 | 200 | 735.458348ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:55 | 200 | 669.741029ms | 192.168.3.10 | POST /v1/vision/detection
 
M

maximosm

Young grasshopper
Joined
Jan 8, 2015
Messages
95
Reaction score
6
Tinbum said:
4GB


[GIN] 2020/12/19 - 14:54:48 | 200 | 622.795749ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:54:56 | 200 | 467.233481ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:01 | 200 | 603.681984ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:04 | 200 | 619.815421ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:12 | 200 | 982.65425ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:15 | 200 | 977.080009ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:18 | 200 | 614.320557ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:52 | 200 | 615.467718ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:55 | 200 | 687.367886ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:55:58 | 200 | 664.312052ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:11 | 200 | 690.881715ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:14 | 200 | 623.645039ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:18 | 200 | 637.22545ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:21 | 200 | 619.848391ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:24 | 200 | 682.57585ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:27 | 200 | 701.910141ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:31 | 200 | 897.778893ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:33 | 200 | 599.392303ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:36 | 200 | 688.946672ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:49 | 200 | 662.386643ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:52 | 200 | 735.458348ms | 192.168.3.10 | POST /v1/vision/detection

[GIN] 2020/12/19 - 14:56:55 | 200 | 669.741029ms | 192.168.3.10 | POST /v1/vision/detection
Click to expand...
Realy realy good.this numbers have the i7 10700 windows that working together with the BI.
 
balucanb

balucanb

Getting the hang of it
Joined
Sep 19, 2020
Messages
147
Reaction score
23
Location
TX
Village Guy said:
What's the state of play? Have you got it recognising any of your own subjects?
Click to expand...
For anyone following along- it most likely will not be hard for you to do- This is my first foray trying to use Docker, cutting and pasting code (No way in hell could I write any) using Portainer, WSL, WSL2, ect.

Village Guy- NO. Just woke up LOL it was (is) kicking my butt, I had to give up last night I was pretty frustrated. I see John has sent me a couple replies that I need to look at. I literally am 1 step from completion. I have the stinking model built, but the last step in deploying it is what has me stumped. I'm on holiday for the next 3 weeks so it will get figured out or else I am going to have a very large electronics fire. LOL.
 
T

Tinbum

Pulling my weight
Joined
Sep 5, 2017
Messages
448
Reaction score
126
Location
UK
You must log in or register to reply here.
Top