Opencv Yolo In_windows

OpenCV, Darknet, Yolov4 in Windows10

https://www.youtube.com/watch?v=sUxAVpzZ8hU

Prerequisition

1. install Visual Studio (with English pack!!)

• C#
• C, C++
• Add English in language pack !!!!

  2. install vcpkg

  install git

  https://git-scm.com/download/win

  install cmake

  https://cmake.org/download/

• check “for all users”

  install vcpkg

  https://vcpkg.io/en/getting-started.html?platform=windows

   $ .bootstrap-vcpkg.bat
   $ .vcpkg integrate install
   $ .vcpkg install <package-name>
  

  https://github.com/microsoft/vcpkg/blob/master/docs/specifications/feature-packages.md

setup system environment

• register name: VCPKG_DEFAULT_TRIPLET , value: x64-windows
• register name: VCPKG_ROOT, value: C:\vcpkg

3. install python3.8

• checkt “add path” (important!!)

  4. install CUDA Toolkit

  install CuDnn

  copy CuDnn files to CuDA

• copy cudnn/cuda/bin/cudnn64_8.dll to program files/nvidia toolkit/cuda/v11.3/bin
• copy cudnn/cuda/include/cudnn.h to program files/nvidia toolkit/cuda/v11.3/include
• copy cudnn/cuda/lib/x64/cudnn.lib to program files/nvidia toolkit/cuda/v11.3/lib/x64

  install vcpkg

  install boost

  $ vcpkg install pthreds
  $ vcpkg install curl
  $ vcpkg install openssl
  $ vcpkg install zlib
  $ vcpkg install boost
  

  install etc

  ``` $ vcpkg install protobuf

• protobuf-c(only supports Linux) $ vcpkg install nats (no package, direct build!)

  ### install opencv
  https://github.com/microsoft/vcpkg/blob/master/docs/specifications/feature-packages.md
  

  $ vcpkg install eigen3 zstr glog $ vcpkg install ffmpeg giflib hdf5 leptonica libjpeg-turbo liblzma libpng libwebp tiff $ vcpkg install opencv $ vcpkg install darknet[full]:x64-windows

  $ vcpkg install darknet[full] –triplet x64-windows

  ### install opencv-python
  

  $ pip install opencv-python

  ### check opencv in python
   run python in cmd
  

  import cv2 cv2.version … ```

install OpenCV (build)

download and setup for cmake

• download opencv source, https://opencv.org/releases/
• download opencv_contrib, https://github.com/opencv/opencv_contrib
• extract two sources
• make opencv folder and move the extracted folders to it

  generate with Cmake

• check BUILD_opencv_world

  build with visual studio

• open ./opencv/build/ALL_BUILD.vcxproj
• set Release / x64 (if 64bit)
• build CMakeTargets/ALL_BUILD
• build CMakeTargets/INSTALL

install Darknet

prepare build

• make folder

   $ cd c:\
   $ mkdir yolov4
  

• download and decompress (C:\yolo_v4\darknet) https://github.com/AlexeyAB/darknet
• copy opencv files to darknet folder

   $ cp /opencv/build/bin/Release/opencv_world452.dll C:\yolo_v4\darknet\build\darknet\x64
   $ cp /opencv/build/bin/Release/opencv_videoio_ffmpeg452_64.dll C:\yolo_v4\darknet\build\darknet\x64
  

• copy cudnn file to darknet folder

   $ cp C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.3\bin\cudnn64_8.dll C:\yolo_v4\darknet\build\darknet\x64
  

  change CUDA version

• cd C:\yolo_v4\darknet\build\darknet
• modify darknet.vcxproj, yolo_cpp_dll.vcxproj
• change CUDA version darknet-

  open yolo_cpp_dll.vcxproj and build it