search

toefel18 / visual-control

Face, eyes, nose, mouth detection and Face Recognition Using Eigenfaces
http://www.youtube.com/watch?v=yFSPnu6_TZY
10 stars 5 forks source link

readme

visual-control

Automatically exported from code.google.com/p/visual-control

Original Idea: By detecting objects in the input of a video capturing device and interpreting their meaning, we try to construct commands. The commands can be used to control anything you like.

To start off, we show head tracking trough face detection with a web cam. Next we will try to recognize familiar persons (using the eigenfaces technique) and fetch their relative information (as students, it will fetch their schedules etc.) Depending on time, we might also implement gesture detection so they can visually control the information they want to see. The system is supposed to be trained with the images of persons in order to recognize them, after that, it should be work automatically whenever a person appears.

This project implements: face detection and face recognition in CPP, together with very simple training functionality.

This project uses C++ (and possibly some C++0x features) in combination with the following libraries:

Boost 1.45.0 OpenCV 2.1 (build with TBB support) Qt 4.7.1

summary Compiling the project and its dependencies

To compile this project, you will need to setup your system. Currently we are using C++ with Visual Studio 2010.

@ Details

Steps:

  1. Install visual studio 2010
  2. Download and compile the boost libraries (I am using version 1.45.0)
    • download boost sources and extract
    • open the visual studio 2010 command line tool and cd to boost root
    • run the 'bootstrap' command to build BJAM
    • run 'bjam variant=debug,release threading=multi link=static'
    • set include and library paths in the visual studio 2010 property pages
  3. Download and compile the OpenCV 2.1 library (2.2 has a bug on win7 with web-cam!)
    • Download CMake 2.8
    • Download threading building blocks (I am using TBB 3.0 Update 4)
    • Download OpenCV 2.1 sources
    • Open CMake-gui and set the source code path to OpenCV root, and where to build binaries to OpenCV root/build
    • run configure (select visual studio 2010 as environment, use native compilers)
    • select with TBB, run configure again
    • search for TBB_PATH_NOT_FOUND, click on that line and enter the path manually
    • press configure again and then click generate
    • go to the OpenCV root/build and open OpenCV.sln, inside the .sln build the debug and release versions.
    • if you get errors concerning TBB, add the include path again in the property pages
  4. Download and compile Qt (I am using 4.7.1, we downloaded the vs2008 windows installer package)
    • open the visual studio 2010 command prompt
    • cd to the Qt root directory and run configure --help. Add some changes if you like
    • make sure you have opengl enabled, use the -mp flag to use multiple cores during compilation, we build the static library variant. this is the build configure command I used: 'configure -platform win32-msvc2010 -static -debug-and-release -mp -no-qt3support -no-webkit -qt-sql-sqlite -plugin-sql-sqlite -fast'
    • when configure finishes, run nmake (this takes 1 to 2 hours)
    • download and install the QT visual studio plugin (i am using 1.1.7)
    • configure visual studio 2010 to use the default qt version
  5. (optional) Install AnkhSVN to access subversion trough Visual Studio
  6. Download the code in the trunk, set the correct include paths and libraries needed (cv210.lib, highgui210.lib, cxcore210.lib) or debug versions (cv210d.lib, highgui210d.lib, cxcore210d.lib). Boost uses auto-linking and QT configures the project trough the vs plugin tool, so no configuration needed for those libraries.
  7. compile
  8. place the dll files in the Release folder next to the executable (cv210.dll, highgui210.dll, cxcore210.dll, tbb.dll) or debug versions in the Debug folder (cv210d.dll, highgui210d.dll, cxcore210d.dll, tbb_debug.dll)
  9. run the executable and enjoy!

Video Tutorials

For compiling Boost, you can also watch this video tutorial: http://www.youtube.com/watch?v=5AmwIwedTCM for Compiling and configuring Boost with Visual Studio 2010

For compiling OpenCV, you can also watch this video tutorial: http://www.youtube.com/watch?v=XeBhwbRoKvk for Compiling and configuring the OpenCV library with Visual Studio 2010

View the demonstration video http://www.youtube.com/watch?v=yFSPnu6_TZY