Recognizer configuration on CentOS

I usually forget some things.

Today, I reinstall my centos system and configure recognizer on it (Recognizer is my online version demo, powered by OpenSSE v1.10. You can get an offline version on my Github). This work actually took me two hours, it’s unbearable thing! So, I decide to record.

Until now, OpenSSE only dependent on OpenCV and boost (Core engine has canceled Qt library in new version). We start from OpenCV.

OpenCV install

  • Install all the required packages using yum
yum groupinstall "Development Tools"
yum install gcc
yum install cmake
yum install git
yum install gtk2-devel
yum install pkgconfig
yum install numpy 
yum install ffmpeg
  • Create working directory and check out the source code
mkdir /opt/working
cd /opt/working
git clone
cd opencv
git checkout tags/

You can chooce any OpenCV version not less than 2.4.

  • Create the Makefile
mkdir release
cd release
  • Build and install
cd /opt/working/opencv/release
make install
  • Add OpenCV to system
cd /etc/
echo "/usr/local/lib/" > opencv.conf
sudo ldconfig

Boost install

tar -zxvf boost_1_55_0.tar.gz
./ --prefix=/usr/local
./b2 install --with=all

OpenSSE install

cd opensse/src
mkdir build
cd build
cmake ..
make install

Recognizer (OpenSSE online version) build

Make sure database has a correct path before build recognizer(params.json and recognize.cpp), and then mkdir /tmp/opensse-tmp/ to cache temp image.

Run web api, everything is going well, Do you want to try?