garricklin/mtcnn-light

this repository is the implementation of MTCNN with no framework, Just need opencv and openblas, support linux and windows

MTCNN-light

Introduction

this repository is the implementation of MTCNN with no framework, Just need opencv and openblas.
"Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Neural Networks", implemented with C++，no framework
it is very easy for you to use.
it is can be a part of your project with no framework, like caffe and mxnet.
it is real time for VGA, and you can improve it's runtime.

Time Cost

The average time cost is about 68ms per frame(640,480).The result is generated by testing a camera. mini_size is 40
cpu i5-4590
os windows10 64bit

Dependencies

opencv 2.0+
openblas

##ubuntu

opencv

you can find many tutorials.

openblas

It is very easy to install
1 download the source code from https://github.com/xianyi/OpenBLAS
2 Extract it and type "cd xxx", xxx means the directory
3 type "make"
4 type "make install PREFIX=your_installation_directory"

if you don't have cmake

apt-get install cmake

usage

cd root_directory
vim CMakeLists.txt
change include_directories(the_directory_of_openblas_include_of_yours)
change link_directories(the_directory_of_openblas_lib_of_yours)
save and exit

cmake .
make
./main

for windows

opencv and openblas

there is binary packages of openblas for windows, you just need download it
But you should to be careful, if you download the 64bit ,you need configure
the opencv and vs project environment with 64bit, don't choose x86