Read the statement by Michael Teeuw here.
Facial Recognition - customize your mirror for every user!
-
@sunnysys OK, I figured it out.
Finally managed Facial-Recognition to work, but I’ve got a little issue: I have to stand real close (10 cm) to RPi Camera to recognize me. What do I wrong, and how can it be fixed?Thanks!
-
@sunnysys What equipment and software did you use?
-
@cowboysdude
RPi2 + RPi camera, control with laptop via Putty+Xming. I installed OpenCV 2.4.9 on RPi2.
I wrote a little script in Python and made 20 pictures with RPi2 + RPi camera. Then I used paviro’s MMM-Facial-Recognition-Tools to train with the saved pictures, also in RPi2. (paviro says it can be pretty slow on RPi, but it wasn’t. :) ) After this I gave MMM-Facial-Recognition module to Magic Mirror config.js and the classis to the other modules.
It works, but can recognize my face rarely. Maybe I made the pictures in inappropriate lighting conditions? -
@sunnysys Somewhere in the code is a hardcoded setting of the resolution that is captured (I will add an option for that someday but I am busy with school right now). If you set this to an higher resolution you will not have to stand as close but it will require more processing power :)
Please also note that you should take the pics in a similar lightning environment as your mirror will be! -
@paviro Really thanks paviro, I will try to find it! :)
-
@sunnysys Thank you!!! Sounds straight forward enough!!
Thank you again!!!
-
@sunnysys OH shoot I just sat down to order some stuff for this and realized I didn’t ask you which camera did you get? Did you get the regular or NOIR version?
Do you mind sharing your py script that you used to make the pictures? Much appreciated!
Thanks!
-
@cowboysdude @sunnysys why not use the capture script in the module repo?
-
@paviro When I get the camera and rebuild the mirror I will do that!! Sounds like it may be easier…
-
@cowboysdude I’ve got the 5MP regular version PiCamera. I read somewhere that the new 8MP ones still got some issues. Don’t know what’s the situation with it now.
Yes, I can share you my capturing code but now I’ve tried to capture images with the built in script of paviro’s MMM-Facial-Recognition-Tools and it works fine. No need to use other script anymore. :)
