PartsBasedDetector

This project implements a Parts Based Detector in C++, described in the following paper:

Yi Yang, Deva Ramanan, "Articulated Pose Estimation with
Flexible Mixtures-of-Parts," CVPR 2011

Setting environment

Installing Octave

$sudo apt-get install libgraphicsmagick++-dev
$wget ftp.gnu.org/gnu/octave/octave-3.8.2.tar.gz
$tar xvf octave-3.8.2.tar.gz
$./configure
$make -j4
$sudo make install

Installing packages

$octave
> pkg install control-2.8.2.tar.gz
> pkg install general-1.3.4.tar.gz
> pkg install signal-1.3.2.tar.gz
> pkg install image-2.2.2.tar.gz

The project has the following dependencies:

• Octave REQUIRED (for learning)
• OpenCV REQUIRED (for image processing)
• CMake REQUIRED (for building)
• Doxygen OPTIONAL (for documentation)
• OpenMP OPTIONAL (for multithreading)
• ROS OPTIONAL (for publishing detections on a ROS topic)
• ECTO OPTIONAL (for building and ECTO cell)

Downloading dataset

$cd <PROJECT_ROOT>/pose/dataset
$./load_inria.sh
$octave load_fashionista.m

Building

The project can be built in one of two modes:

• A standalone binary (for testing functionality)
• A shared library (for use in existing applications)

To configure the project, set the options at the top of CMakeLists.txt To build the project, follow the normal cmake routine from the root folder:

git clone git@github.com:IshitaTakeshi/PartsBasedDetector.git --recursive
cd PartsBasedDetector/cvmatio
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=.. ..
make -j4
make install
cd ../..
mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=.. ..
make -j4
make install

Learning

The learning code is currently only in Octave/Matlab. This is because the detector supports a number of learning schema, and porting all of these to C++ is not practical at this time. Please consult the README within the matlab/ directory for instructions on training a model

This package is developed and maintained by Hilton Bristow, Willow Garage

Generating models

$cd matlab
$octave compile.m
$octave demo_fashionista.m

Detecting

cd build
./src/PartsBasedDetector ../matlab/fashionista.mat <path to image>