Open Detection
1.0
|
The compilation and installation of OD is fairly simple. You first have to install its dependencies before compiling. Folllowing are the dependencies with their specific settings:
OpenCV 3.0 is to be compiled with the modules xfeatures2d (for features like SIFT) and CUDA.
OpenCV contrib - for xfeatures2d module handing SIFT/SURF features
Detailed instructions with source are provided here: https://github.com/itseez/opencv_contrib . You need to download this seperate repository before compiling OpenCV.
CMAKE options: OPENCV_EXTRA_MODULES_PATH=<path_to_opencv_contrib>/modules
OpenCV CUDA module - for GPU enabled feature detectors and matcher.
CMAKE options: WITH_CUDA=ON
Download and compile latest VTK with its default settings.
3d_rec_framework - for ESF, ESF etc recognition pipeline. To build with this setting you need to install OpenNI as well, which is the mandatory dependency for this app. Please refer to the PCL website for the version information.
Additional CMAKE options: BUILD_apps=ON, BUILD_apps_3d_rec_framework=ON
Get the latest version of Eigen (source) using your package manager.
We use SVMLight for training. To user the feature of training using SVMLight, you need to download the source and put in a specific folder (we could not include the source in 3rdparty because of its restrictive licence). Please see http://svmlight.joachims.org/ for the details and terms of use.
Download the source code from http://svmlight.joachims.org/ and extract it to opendetection/3rdparty/svmlight
. You should have files like svm_common.c
, svm_learn.c
, etc. under the directory opendetection/3rdparty/svmlight
.
With the above dependencies installed, OD should compile without any problem. Download the source from https://github.com/krips89/opendetection and compile it with default cmake options. The code while platform independent, is only tested and run in Linux machine. Instructions for the usage for linux are provided below:
Compile out of source using cmake+your favorite compiler. For example:
Download the code:
configure with CMake and compile