5.4.1 - April 3, 2021 - Support for DeepStack facial recognition

StphNieuw

n3wb
Joined
Aug 1, 2020
Messages
15
Reaction score
5
Location
Nola
Any one have any luck with the facial detection after all of the updates. My object detection is great, just the facial recognition doesn't appear to work. Have even taken the 'new faces' photos and uploaded them with no effect.
 

m_listed

Getting the hang of it
Joined
Jun 11, 2016
Messages
176
Reaction score
57
Any one have any luck with the facial detection after all of the updates. My object detection is great, just the facial recognition doesn't appear to work. Have even taken the 'new faces' photos and uploaded them with no effect.
You have to put the face names into the “To confirm” box in Trigger → AI settings for each camera. Also put “unknown” in there too for it to note any unknown faces.
 

anijet

Pulling my weight
Joined
Mar 10, 2014
Messages
346
Reaction score
165
Location
CA
Why do this: "put the face names into the “To confirm” box in Trigger → AI settings for each camera"?
Just put unknown in the "To confirm". Works great daytime and OK at night for me.
 

Attachments

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
24,432
Reaction score
47,556
Location
USA
You would put face names in to either record or not record or do some other function based on it recognizing a face you want. But if that isn't what you are looking to do, then don't put names to faces and just go with people LOL.
 

StphNieuw

n3wb
Joined
Aug 1, 2020
Messages
15
Reaction score
5
Location
Nola
To confirm based on the comments, if I want AI to detect person and cancel the alert based on facial detection I would put:

To Confirm: person, unknown
To Cancel: Stephan,Stephan1, Stephan2, Stephan3 etc.

In the overall BI AI settings I have corresponding pictures uploaded with those exact names.

Or did I mess it up :)
 

Attachments

m_listed

Getting the hang of it
Joined
Jun 11, 2016
Messages
176
Reaction score
57
To confirm based on the comments, if I want AI to detect person and cancel the alert based on facial detection I would put:

To Confirm: person, unknown
To Cancel: Stephan,Stephan1, Stephan2, Stephan3 etc.

In the overall BI AI settings I have corresponding pictures uploaded with those exact names.

Or did I mess it up :)
Looks good, except I don’t know if it allows spaces after commas. Also, having things both in the “To confirm” and “To cancel” boxes stops all detection for me. Is it working for you?
 

StphNieuw

n3wb
Joined
Aug 1, 2020
Messages
15
Reaction score
5
Location
Nola
Hi all, I'm still struggling with this and a little perplexed at this point. I get the alerts 90-100% of the time correctly with the correct object designation but cannot get facial recognition to stop triggering the alert.

I tried taking out the spaces between the commas out as suggested but that didn't appear to make a difference.

I went ahead and deleted the To Confirm entries and left the entries in the To Cancel (without spaces) as also suggested and didn't have success. I have uploaded multiple images sent to me from the alerts, both full body (with same clothing etc) and faces only didn't seem to make a difference. Facial recognition is checked in the general settings --> AI.
Tried it the otherway around as well without results. Any other suggestions?
 

austwhite

Getting the hang of it
Joined
Jun 7, 2020
Messages
93
Reaction score
92
Location
USA
The facial recognition appears hit and miss. Does it use the substream instead of high res jpgs?
It would be nice to group multiple photos under one label so I didnt have to have 10 or more photos of myself just so it recognises me. The DS documentation is a bit vague, but seems to allow for this in DS.
 

StphNieuw

n3wb
Joined
Aug 1, 2020
Messages
15
Reaction score
5
Location
Nola
I have been using a single hi-res feed @ 8mp and cameras at slightly above face height with good visibility. I am continuing to tinker with it to see what it will work.
 

BORIStheBLADE

Getting comfortable
Joined
Feb 14, 2016
Messages
739
Reaction score
2,065
Location
North Texas
The facial recognition appears hit and miss. Does it use the substream instead of high res jpgs?
It would be nice to group multiple photos under one label so I didnt have to have 10 or more photos of myself just so it recognises me. The DS documentation is a bit vague, but seems to allow for this in DS.
I believe it uses high resolution if you have recordings set to continuous. The new setting "continuous + alerts" I'm under the impression uses the substream for Deepstack.
 
Top