Facial recognition

[sdetweil]

@majik655 things are constantly changing. I’ve only installed the MMM-face-reco-dnn on my Pi4

it’s a lot of work

[majik655]

@sdetweil said in Facial recognition:

@majik655 things are constantly changing. I’ve only installed the MMM-face-reco-dnn on my Pi4

it’s a lot of work

Yeah iv’e been through the process and back with the very beginning of face recog I was able to get it done but that was a long time ago… it seems it has come a long way.

I guess I get confused when it comes to the dataset… can I just put my own pictures from say my phone in there? Or do I have to somehow figure out how to take pictures and such with the camera on the pi?

The links in the instructions say “OpenCV 4.1.2 quick installation (wthout compliation) for Raspbian Buster”
Bu takes you to a page that is 4.3… for the opencv part is that the only part that is needed is this:
https://github.com/cyysky/OpenCV-Raspberry-Pi-4-Package-for-Python

Anyway made a backup of my mirror…now lets see if I can destroy it trying to get this working.
Thank you for your replies!

[sdetweil]

@majik655 I took pics w my phone

[sdetweil]

@majik655 instructions written when opencv 4.1 was current, now 4.3. fun times

[majik655]

@sdetweil
Ok thats what I was hoping for …we will see how far I get lol Thanks again (I might have another question for ya but I will try not to)
Thanks

[majik655]

@sdetweil Crap see another question… how many pics did you load in the beginning from your phone?

[sdetweil]

@majik655 I used 4 pics

[majik655]

@sdetweil said in Facial recognition:

@majik655 I used 4 pics

lol well I just uploaded 17 and im wondering if my PIR sensor or other module configs are stopping this …
I havent figured out yet if the camera is even working good… I am using a usb cam so I have it set to 0 in the config…

But I have a motion sensor, mmm-pages and google assistant and assistant2display which I wonder is causing any issues.

The mirror boots up loads modules and then goes blank because I have all modules set to me (no default or everyone).

I need to figure out how to view live camera… I wish there was a way for the module to test or report what it sees or doesnt see for that matter.

[majik655]

@sdetweil
I finally got it!
I commented out basically everything in my config except for the basics like clock and such…
Nothing was working …
Then I realized the lights on the camera were not turning on when the MM started…which means there was no way it was going to recognize anything if it wasnt even on lol

To get it to work I had to change in the module config

source: 0,

to

source: '/dev/video0',

Which by the info source: 0, SHOULD BE /dev/Video0

Not sure why 0 didnt work but it works now with /dev/Video0!!!

Thank you for listening!

[sdetweil]

@majik655 nice work!

[majik655]

@sdetweil
Yeah so one thing I have found is it is kind of worthless to have a motion sensor running that hides the modules after 2minutes of no motion…because if you block the sensor once it has closed… it looks like the camera cant turn it back on because the sensor has hidden everything.

I either turn off the motion sensor and rely on the camera all the time …
or
somehow have the motion detection turn on the camera recognition part…recognize then show correct modules.
or
have the recognition…trigger the motion sensor

so many decisions I dont know which to do lol

[Alphons]

I wonder how long it takes to encode 5 pics with a rpi4? Encode.py is running for 3 hours now and i hope ther will bee a “Oke ready” thingy.

[sdetweil]

@Alphons should take like 30 seconds

[Alphons]

@sdetweil and no “ready” after serializing encodings… message?

[sdetweil]

@Alphons I don’t remember. I think the encode script just ended when is was done

[Alphons]

@sdetweil oke, then i did something wrong. Started the encode.py, it sorted the pics en encodes them…for 4 hours now :)