================================================================================
What's in this package ??

Torch 5          Torch5 provides a Matlab-like environment for state-of-the-art 
                 machine learning algorithms. It is easy to use and provides a 
                 very efficient implementation, thanks to an easy and fast 
                 scripting language (Lua) and a underlying C implementation.
                 
                 The Torch 5 library is re-distributed here for simplicity of 
                 installation.
                 The original package can be found here:
                 http://torch5.sourceforge.net
              
                 The distribution has been slightly modified, in particular, 
                 the original Lua kernel has been patched for multi-threaded 
                 applications.
                 
                 Torch is licensed under a BSD license:
                 http://torch5.sourceforge.net/manual/License.html
              
xLearn           xLearn is an extension library for torch. It provides dozens 
                 of tools/modules for vision, image processing, and machine 
                 learning for vision
              
luaFlow          luaFlow is a unified flow-graph description environment for
[beta]           vision / image-processing types of applications. One of its
                 primary objectives is to abstract computing platforms, by 
                 providing a unified, high-level description flow.
              
xFlow            a serializing language for luaFlow, that allows algorithms to
[beta]           be imported/exported from/to other software frameworks
              
neuFlow          neuFlow is the compiler toolkit for the neuFlow processor, 
                 developped at New York University / Yale University.
                 The neuFlow processor is dataflow computer optimized for
                 vision and bio-inspired models of vision. 
                 The neuFlow compiler currently converts xLearn/torch algorithms
                 to native neuFlow's bytecode.
                 Soon to appear is a luaFlow>neuFlow compiler, which would   
                 simplify retargetting.
                 It is quite important to have access to a neuFlow device to 
                 be able to experiment with it: for more info/support, to get
                 a neuFlow-enabled board, please contact clement.farabet@gmail.com
              
thread           the Lua core is patched with LuaThread to allow
                 multithreaded apps
              
LuaJIT           the entire framework can be built against LuaJIT for
                 improved performance
              
opencv           a wrapper for OpenCV, for now just a couple of functions, 
                 super easy to extend
              
debugger         the open-source debugger framework for Lua (activated by 
                 requiring 'debug')
              
camiface         a wrapper for libcamiface, to interface webcams on MacOS
              
video4linux      a wrapper for libv4l2, to interface webcams in Linux

mstsegm          a wrapper around P. Felzenszwalb’s image segmentation lib

powerwatersegm   a wrapper around C. Couprie’s Powerwatershed lib

stereo           a wrapper around P. Felzenszwalb’s BP-based stereo code

opticalFlow      a wrapper around C. Liu’s great optical-flow estimator

kinect           a wrapper around Microsoft's kinect device

pink             a wrapper around M. Couprie's Morphology library


================================================================================
INSTALL

(1-LINUX) 
    install dependencies (compilation tools, cmake, QT4):
    $ sudo apt-get install binutils gcc g++ cmake libqt4-core libqt4-dev libqt4-gui libreadline5-dev libpcap-dev

    optionally, install OpenCV 2.x, to get access to extra packages:
    http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/2.1/

(1-MACOS)
    install dependencies (readline, cmake, QT4)
    $ sudo port install readline-5 cmake qt4-mac-devel

    you might want to use a prebuilt version of QT4, to avoid the 2 hour
    build time... (I still don't understand
    why MacPort relies on sources rather than binaries...)
    I'm keeping a working version here (installs in 5mins):
    http://data.clement.farabet.net/qt/qt-mac-cocoa-opensource-4.5.3.dmg

    optionally, install OpenCV 2.x, to get access to extra packages
    on Snow Leopard, that works: 
    sudo port install opencv +sl_64bit

(2-COMMON) 
    once the dependencies are installed, just run:
    $ make
    $ [sudo] make install
    for the default install

    or just
    $ make help
    for more info about the options/submodules

    example of a local install:
    $ make install INSTALL_PREFIX=~/local


================================================================================
DOC

The documentation is built and installed, and then available at:
/usr/local/share/torch/html/index.html
or
INSTALL_PREFIX/share/torch/html/index.html

All details about installation are available there.