Ubuntu系统编译安装SeetaFace

深度学习交流QQ群：116270156

一.安装OpenCV 依赖

1. Require Packages

[compiler] sudo apt-get install build-essential
[required] sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
[optional] sudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev

2. Get OpenCV Source Code
http://opencv.org/downloads.html

3. Building OpenCV from Source code Using Cmake & Command Line

cd ~/opencv #移动到源文件文件夹中
mkdir release
cd release
cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local ..


#在上面步骤结束之后进行make
make 
sudo make install #该命令行必须

4. Test OpenCV

cd opencv_samples
./build_all_sh
.facedetect  Pathofimage

二.安装SeetaFace Detection

下载源码

git clone https://github.com/seetaface/SeetaFaceEngine.git

build on linux

mkdir build #在FaceDetection目录中  
cd build  
cmake ..  
make -j${npoc}

可能出现/usr/bin/ld: cannot find -lopencv_dep_cudart
解决方法，在执行cmake的时候加入下面的参数CUDA_USE_STATIC_CUDA_RUNTIME=OFF，例如cmake -D CUDA_USE_STATIC_CUDA_RUNTIME=OFF ..

run demo

./build/facedet_test imagefilePath ./model/seeta_fd_frontal_V1.0.bin

error: ‘isnan’ was not declared in this scope
解决方法，修改文件中的isnan为“std::isnan”

三.安装SeetaFace Alignment

Build in Linux

 mkdir build #在FaceAlignment目录中

然后将FaceDetection中的/include/face_detection.h和/build/libseeta_facedet_lib.so 拷贝到/FaceAlignment/build文件夹下，为了生成test example

cd build  
cmake .. 
make

如果一切正常，可以用测试程序来测试一下，先要拷贝/FaceDetection/model文件夹下的seeta_fd_frontal_v1.0.bin文件到FaceAlignment的build中

./fa_test

用到测试程序的时候可能出现以下问题
iofread.c没有这个目录或文件
该问题在于生成fa_test中的face_alignment_test.cpp文件中的路径有些问题比如/data和/model文件夹,还有需要用到的图片路径

四.安装SeetaFace Indentification

build on linux

    mkdir build
    cd build
    cmake .. && make

注意以下问题

未定义引用对于之前生成的FaceDetection和FaceAlignment库中定义的函数。
对于前面FaceDetection和FaceAlignment build生成的两个库的libseeta_facedet_lib,libseeta_fa_lib添加到FaceIdentification/build文件夹中。
然后修改src/test/CMakeLists.txt文件，在Cmakelists文件中添加查找的链接库link_directories(${PROJECT_BINARY_DIR})(也就是在当前build文件夹下查找，该命令最好添加循环函数外边)
在循环中补全target_link_libraries(XXXXX seeta_facedet_lib seeta_fa_lib)

#`src/test/CMakeLists.txt`修改之后如下
aux_source_directory (. SRC_LIST)

message(${SRC_LIST})
link_directories(${PROJECT_BINARY_DIR}) #添加lib
# add external libraries
find_package(OpenCV REQUIRED)

enable_testing ()
foreach (f ${SRC_LIST})
  string(REGEX REPLACE "[.]cpp" ".bin" BIN ${f})
  add_executable(${BIN} ${f})
  target_link_libraries(${BIN} viplnet ${OpenCV_LIBS}  seeta_facedet_lib seeta_fa_lib) #添加lib
endforeach ()

2.model/seeta_fr_v1.0.bin not found ：解决解压缩model中的两个文件
3.face_alignment.h, face_detection.h: No such file or directory,出现这个的原因是sample在编译是需要用到这些头文件。解决方法，将SeetaFaceEngine/FaceDetection/include/face_detection.h和SeetaFaceEngine/FaceAlignment/include/face_alignment.h拷贝到SeetaFaceEngine/FaceIdentification/include/文件夹中

4.分别修改test_face_recognizer.cpp和test_face_verification.cpp中data和model的相对路径，还要修改test_face_verification.cpp文件中查找之前编译过得模块的路径

Run demo

    ./build/src/test/test_face_recognizer.bin


    ./build/src/test/test_face_verification.bin #比较两个文件相似度

谷歌学术教程