Question on dlib component

Hi

Have you test tout do that before ?
$sudo apt-get install build-essential cmake
$ sudo apt-get install libgtk-3-dev
$ sudo apt-get install libboost-all-dev

Hiā€¦ Thanks for suggestingā€¦ Installed those 3 and tried to compile cmake which went through. However, installation of face_recognition failed and HA started hanging / crashed.

Does it really work in a Raspberry PI3 All-in-One HA installation.?

On a Rpi3 i donā€™t know, for me HA is installed on a VM on a ESX serverā€¦

Additional dependencies:
sudo apt-get install libjpeg-dev zlib1g-dev liblapack-dev

Hi all,
for those who made this component work. How well does it work? How high is your accuracy ?
Also, will it be able to do inference on PI? Or need better CPU to run on?

Where did you do model training if any?

How does this component compare to Microsoft Face recognition API? Their model must be better, but I wonder what the latency is for dlib vs Microsoft Face api?

Hi,

I tested that one year agoā€¦
The Microsoft solution is connected to microsoft server i prefer avoid this type of architecture.
Nevertheless this solution is quite quickly.

I test DLIB component and on a correct server (XEON 3.9ghz) on a VM, it takes 4-5 seconds to recognize somebody and execute an automation.

For me itā€™s not enough quick to be integrated to my systemā€¦ but it is fun :smile:

Have no idea how to get it work on Hassio on RPI3.
After container restarting installed packages (cmake and face_recognition) gone.

If you have a configuration that works on a normal install (not container based) and are willing to share that, I will figure out the container resolution. It maybe just mapping some files thatā€™s required, similar to the t solution for Zwave configuration

Regards
Damian

Heyā€¦ i wonder if the configuration options keep_images_with faces and so on have it ever made into the master branch? I do not see those configuration options in the dlib documentation - why is that?

Also i tried to use this component on a Raspberry Pi3 but it slowed down the hole system - but the strange thing was that it did not take CPU or memory resources, so i donā€™t really understand why it almost blocked my Raspberry. This should not be an IO intensive task, right?
Has anyone achived to got this smoothly running on an RPi3 with camera.local_file as source?

Hi,
I have the same question as tom-x1 :slight_smile:

And I dā€™ont understand where the ā€œface.jpgā€ are