• Hans-Jörg posted an update 4 years, 3 months ago

    Hi,

    I start to build my InMoov with the head, because it is the best start where you can see a result quickly.
    I used the modifed head with rasperry pi 3 and arduino uno inside the head, so I have only one cable for the Batterie outside. The Problem: I´m not able to get the USB-Soundcard working with MRL.
    , so I have no microphone.
    Yesterday I conect the components with a Nettop-PC with Windows 10, It works fine, but when I start “search human” ( MRL 1.0.1412 32Bit, InMoov3.full.byGael.Langevin.1.py ) my InMoov eyes fixed my face, but I can´t stop it with earcomands, Google Chrome gets no Power from the CPU.

    OK, a Netbook with 1,8GHz 32Bit ist not the fastest, so I tried my biggest PC with 4×3,5GHz and Windows 10 64Bit. I used MRL1.0.1412 64Bit and whats happend? Error, openCV does not work correctly so I tried MRL1.0.1412 32 Bit.
    InMoov starts, I say “search human” and whats happend? The same as with the Nettop the eyes move and Google Chrome got not enough Power to work. The only Different between the smal and the big PC is the number of Frames per second and the Speed of the eye movement.

    What ist wrong?

    • Hi.
      Did you set your microphone into Chrome settings? Make sure your mic is correctly selected and not set for the default. I had the similar issue, once the servo’s would be activated, it seemed Chrome was not listening anymore. It’s important the servos can generate a lot of noise.
      https://www.google.com/chrome/browser/desktop/index.html

    • Hi Gael,

      thank you for your fast Response!

      The Sound-Jack from the Raspi 3 is for Audio/Video Output only!
      In Chrome under Windows the microphone is enabled and without camera/facedetect all ear.commands works without delay.
      When I start the ear.command “search humans” the camera starts with 15 frames/second and the eyes move correct, but it needs 99% of the Power from the CPU. It Looks like, the Chrome lost the Connection. Sometimes I have to update the localhost-page before I can try again voice recognition.

      For Explanation:

      I connect to Arduino Mega2650 with my PC ( leftport/rightport), only the eyeX,eyeY and the mouth-Servo are connectet with the board. The headtracking is disabled.

    • Ups, two Arduino-Boards are connected.

    • Today I made a new Installation from myrobotlab_X64 on WIN10 and it works!
      I didn´t know why, but opencv works and with facedetect the Programm needs 67% off the CPU Power only (4×3,5GHz).

      • Good news, just to make sure, before restarting MRL, you need to kill all the java and javaw process in task manager. In my ZIP folder I have added a .batch file which will kill all these process for you at once.
        Face tracking is a CPU eater with the recent version of OpenCV. I do not know why, but with the older version of MRL, 1.0.107, OpenCV was not so hungry.
        Concerning the audio input, you are correct, my search on the internet got me to the same conclusion.

        • Hi Gael,

          the second positiv message, myrobotlab.1.0.1412 (32Bit) and the Raspberry Pi3 works too!. I get 5 up to 6 fps with facedetect with 65% Power of the CPU.
          I start the WebGui on a second PC with Microphone.

          Thanks for your Time!