CentOS 7 + NVIDIA Driver + CMake3.17 + CUDA10.0 + CUDNN 7.6.5 + FFmpeg + OpenCV4.2.0 + DarkNet

识别算法CentOS部署手册

这是一个目标检测、人脸识别项目的部署手册，包含了一些开发环境的配置方法，相关包已经上传到了我的资源。

环境需求

• CentOS 7
• NVIDIA Driver
• CMake 3.17
• CUDA 10.0
• CUDNN 7.6.5
• FFmpeg
• OpenCV 4.2.0
• DarkNet
• SeetaFace6

安装CentOS

• 正常安装CentOS 7

安装NVIDIA显卡驱动

• 驱动下载
  • https://www.nvidia.cn/Download/index.aspx?lang=cn#
  • 上述地址下载所需显卡驱动
• 环境准备
  • 环境依赖
    • 安装kernel-devel
      • sudo yum install kernel-devel
    • 检查内核版本和源码版本，保证一致
      • ls /boot | grep vmlinuz
      • rpm -aq | grep kernel-devel
    • 安装gcc
      • sudo yum install gcc
    • 安装make
      • sudo yum install make
    • 禁用nouveau
      • sudo vim /etc/modprobe.d/blacklist-nouveau.conf
      • 在 blacklist-nouveau.conf 中插入如下两行
        blacklist nouveau
options nouveau modeset=0
    • 重建initramfs image
      • mv /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r).img.bak
      • dracut /boot/initramfs-$(uname -r).img $(uname -r)
      • reboot
    • 验证是否生效，无输出则生效
      • lsmod | grep nouveau
    • 修改运行级别为文本模式
      • sudo systemctl set-default multi-user.target
      • reboot
• 安装
  • 运行安装程序
    • sudo sh NVIDIA-XXXX-XXXX.run
    • install nvidia’s 32-bit
      • no
    • 其他选项
      • yes
  • 验证安装
    • nvidia-smi

安装CMake 3.17.5

• 下载
  • https://cmake.org/download/
  • 上述地址下载 CMake 3.17.5
• 安装
  • 解压压缩包
  • 配置环境变量
    • vim ~/.bashrc
    • 在bashrc最后添加 export PATH=XXXXXXX/bin:$PATH
    • XXXXXXX 为CMake解压后的路径
    • source .bashrc
• 验证
  • cmake --version
  • cmake-gui --version

安装CUDA和CUDNN

• 下载
  • https://developer.nvidia.com/cuda-10.0-download-archive
  • 上述地址下载cuda10.0及其补丁包
  • https://developer.nvidia.com/cudnn
  • 上述地址下载对应cudnn
• 安装CUDA
  • sudo ./cuda_10.0.130_410.48_linux.run
    • Do you accept the previously read EULA?
      • accept
    • Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 410.48?
      • no
    • Install the CUDA 10.0 Toolkit?
      • yes
    • Enter Toolkit Location [ default is /usr/local/cuda-10.0 ]:
      • 回车确定选择默认位置
    • Do you want to install a symbolic link at /usr/local/cuda?
      • yes
    • Install the CUDA 10.0 Samples?
      • yes
    • Enter CUDA Samples Location [ default is /home/yfzx ]:
      • 回车确定选择默认位置
  • sudo ./cuda_10.0.130.1_linux.run
    • Do you accept the previously read EULA?
      • accept
    • Enter CUDA Toolkit installation directory [ default is /usr/local/cuda-10.0 ]:
      • 回车确定选择默认位置
  • 配置环境变量
    • vim ~/.bashrc
    • 在bashrc后面添加
      export PATH=/usr/local/cuda-10.0/bin:$PATH export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
    • source .bashrc
  • 验证
    • nvcc -V
• 安装CUDNN
  • 解压压缩包
  • cd cudnn-XXXX/cuda/
    • sudo cp include/cudnn.h /usr/local/cuda-10.0/include/
    • sudo cp lib64/libcudnn* /usr/local/cuda-10.0/lib64/
    • sudo chmod a+r /usr/local/cuda-10.0/include/cudnn.h
    • sudo chmod a+r /usr/local/cuda-10.0/lib64/libcudnn*
  • 验证
    • cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
• 修改cudnn的软连接
  • cd /usr/local/cuda-10.0/lib64/
  • sudo rm libcudnn.so.7
  • sudo ln -s libcudnn.so.7.6.5 libcudnn.so.7

编译安装OpenCV4.2.0

• 根据OpenCV4.2.0压缩包完成以下操作
• 环境依赖
  • sudo yum -y install epel-release
  • wget https://download1.rpmfusion.org/free/el/rpmfusion-free-release-7.noarch.rpm 2
  • wget https://download1.rpmfusion.org/nonfree/el/rpmfusion-nonfree-release-7.noarch.rpm 1
  • rpm -ivh *.rpm
  • rpm --import http://li.nux.ro/download/nux/RPM-GPG-KEY-nux.ro
  • rpm -Uvh http://li.nux.ro/download/nux/dextop/el7/x86_64/nux-dextop-release-0-1.el7.nux.noarch.rpm
  • sudo yum install ffmpeg
  • sudo yum install ffmpeg-devel
  • sudo yum install gtk2
  • sudo yum install gtk2-devel
  • sudo yum install libdc1394-devel
  • sudo yum install libv4l-devel
  • sudo yum install gstreamer-plugins-base-devel
  • sudo yum install jsoncpp
  • sudo yum install jsoncpp-devel
• 验证FFmpeg安装成功
  • ffmpeg
• 解压压缩包
  • cd /opencv4.2.0/opencv4.2.0
  • mkdir build
  • cd build
  • cmake ..或者cmake-gui ..
    • CMAKE_INSTALL_PREFIX=/usr/local/opencv4.2
    • OPENCV_EXTRA_MODULES_PATH=XXXX/opencv_contrib-4.2.0/modules opencv_contrib-4.2.0文件在opencv4.2压缩包内，填入路径即可
    • OPENCV_GENERATE_PKGCONFIG=YES
    • WITH_CUDA=ON
    • OPENCV_DNN_CUDA=ON
    • OPENCV_ENABLE_NONFREE=ON
    • CUDA_ARCH_BIN=6.0 6.1 7.0 7.5
    • 确保cmake信息中Video I/O下FFMPEG信息全为YES
  • 上一步的完整命令 XXXXXX替换成安装路径即可 cmake -D CMAKE_INSTALL_PREFIX=/usr/local/opencv4.2 -D OPENCV_EXTRA_MODULES_PATH=XXXXXX/opencv4.2.0/opencv_contrib-4.2.0/modules -D OPENCV_GENERATE_PKGCONFIG=YES -D WITH_CUDA=ON -D OPENCV_DNN_CUDA=ON -D OPENCV_ENABLE_NONFREE=ON -D CUDA_ARCH_BIN=6.0 6.1 7.0 7.5 ..
  • make
  • sudo make install

编译安装DarkNet

• 下载
  • https://github.com/AlexeyAB/darknet
  • 上述地址下载DarkNet
• 安装
  • 解压压缩包
  • 修改makefile文件
    • GPU=1
    • CUDNN=1
    • OPENCV=1
    • LIBSO=1
    • 修改 LDFLAGS中CUDA和CUDNN路径为本机安装路径
    • 修改 COMMON中CUDA路径为本机安装路径
  • cd darknet
  • mkdir build
  • cd build
  • cmake ..
  • make
  • make install
• 将动态库和头文件移动到/usr/local/下
  • cd /usr/local
  • sudo mkdir darknet
  • cd darknet
  • sudo mkdir lib
  • 将编译后的darknet文件夹中的libdrak.so移动到/usr/local/darknet/lib/下
  • 将编译后的darknet文件夹中的include文件夹移动到/usr/local/darknet/下

安装SeetaFace6

• 下载
  • https://github.com/yisampi/SeetaFace6
  • 上述地址下载Ubuntu系统的压缩包
• 安装
  • 解压压缩包
  • 将解压后的文件夹重命名为seetaface6
  • 将seetaface6文件夹移动到/usr/local/下

动态库配置

• cd /etc/ld.so.conf.d
• sudo vim cuda.conf
  • 添加/usr/local/cuda-10.0/lib64
• sudo vim darknet.conf
  • 添加/usr/local/darknet/lib
• sudo vim opencv4.conf
  • 添加/usr/local/opencv4.2/lib64
• sudo vim seetaface6.conf
  • 添加/usr/local/seetaface6/lib64
• sudo ldconfig