FaceDetectionStream

Realtime face detection with ffserver streaming output to local network

This code runs out of the box on this preconfigured virtual machine

VM Configuration:

• 4 cores for the VM is the best setting in my case (corei7 machine).
• Install VMware tools is a must. That can help increase the speed very much.
• Install ffmpeg3 by running this:

  sudo add-apt-repository ppa:jonathonf/ffmpeg-3
  sudo apt update && sudo apt install ffmpeg libav-tools x264 x265

• Install twisted, autobahn, and websocket-client to host a simple echo server

  sudo pip install twisted
  sudo pip install autobahn
  sudo pip install websocket-client

• Install dependecy to run google text-to-speech service, using this gTTS library. To play the mp3 file produced by the gTTS library, we use pyglet that uses AVbin underlying. You can install everything by:

  sudo pip install gtts
  sudo pip install pyglet
  curl -L -s https://github.com/downloads/AVbin/AVbin/install-avbin-linux-x86-64-v10 > installAVbin.sh; sudo sh ./installAVbin.sh
  rm installAVbin.sh

How to run:

cd $FaceDetectionStream
./runVideoDetection.sh

How to connect to the output streams:

The realtime face detection stream and the faceWall can be viewed at:

localhost:8000/stat.html

The echo server that keeps announcing newly detected faces is hosted at:

localhost:8080

To let other machine to access to the services hosted inside VMware, you must set the network of the virtual machine to use VMNet0 interface. That is a bridged interface. Then go to Edit -> Virtual Network Editor -> Change Settings (bottom right) -> Select VMNet0 -> in the Bridge to box, select the network apdater that you want to bridge to (the one that the host machine is using to connect to the local network).

How to change faces and welcome messages

Face examples are stored in images/faces Welcome messages are stored in ./welcomeMessages.txt Names of face examples images and names in welcomeMessages.txt must match