Dlib, why dont you recognize me?

Configuration
1

What am I missing?

The only thing I get is this dlib value.

device_class: face
faces: [object Object]
friendly_name: Dlib Face myface
total_faces: 1

Yea, it somewhat works… It detects faces, but it doesnt recognize them.

image_processing:
 - platform: dlib_face_identify
   source:
    - entity_id: camera.myface
   faces:
     me: /home/homeassistant/.homeassistant/www/faces/me.jpg
2

Alright, I figured out the problem. It was between the chair and the keyboard. That means, me.

The value

faces: [object Object]

when i used the template editor using this template:
{{states.image_processing.dlib_face_myface.attributes}}
it came up with this:
{‘device_class’: ‘face’, ‘faces’: [{‘name’: ‘me’, ‘entity_id’: ‘image_processing.dlib_face_myface’}], ‘friendly_name’: ‘Dlib Face myface’, ‘total_faces’: 1}

And when i used this template:
{{states.image_processing.dlib_face_myface.attributes.faces}}
it came up with this:
[{‘name’: ‘me’, ‘entity_id’: ‘image_processing.dlib_face_myface’}]

Now, I want to filter the name and I cant find the way!
How can I use a template to only show the name?

3

Hi,
I don’t understand how the component can identify my face ?

I have to put a picture of my face in the path me: /home/homeassistant/.homeassistant/www/faces/me.jpg ?

4

That is correct.
Check the component documentation for more info:

5

It can identify a person by learning from just a single photo?

6

Yes. The only thing I did was to put one image of mine and one of my wife’s and it could recognize us both succesfuly. I have removed the component a long time ago but if I remember correctly, it could recognize us even when we had our heads slightly sideways. It gave me the feeling it only needed half the face to do the recognition.

1 Like
7

I’m using Facebox for face recognition at the moment. I’m just curious which one gives the best results. Will try this out soon to compare. Thanks.

8

Hi, do you able to find the way to get the identification on automation. Suppose you get detected and an automated trigger happen. It will help a lot if the identification is accurate.Thanks.

9

dlib needs some updating. I would propose adding the image_processing.face_recognised event similar to that implemented by facebox. @masterkenobi how are you using face recognition for auth?