What is OpenPose

OpenPose is a library that detects the human body (knees, elbows, eyes, etc.) and visualizes the detected results. And this technology was announced at CVPR2017, and the skeleton detection algorithm with a monocular camera using Deep Learning was made into a library and released under the name of OpenPose.

Example:

This chapter mainly introduces how to install and compile OpenPose.

How to install OpenPose

First, https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/doc/installation.md Go to this link, go to Windows and install one by one in the order you wrote them.

(Click here for the version in Japanese and omitted to some extent)

** If you click the blue letters on the ↑ site directly, you can go directly to the corresponding site, so let's install while referring to this **

Let's go to "Windows" on the above site.

** If you click the blue letters on the ↑ site directly, you can go directly to the corresponding site, so let's install while referring to this **

① Installation --The first thing to install in Demo is:

(1)Visual Studio 2015 Install the editor (if you already have one)

(2)CUDA 8 Install in the default location `, C: \ Program Files \ NVIDIA GPU Computing Toolkit \ CUDA \ v8.0``` (3)cuDNN 5.1 Unzip the .Zip, take out only the contents and paste it in `C: \ Program Files \ NVIDIA GPU Computing Toolkit \ CUDA \ v8.0```

2.Installation - Library ② Install the library. (1)Python 2.4.13 64 bits - Windows x86-64 MSI installer. Install in the default location `` `C: \ Python27 (default) . And issue a command pip install numpy protobuf hypothesis```Type the command.

(2) Cmake: After installation, record the path and

http://next.matrix.jp/config-path-win7.html

Let's register the path of the environment variable by referring to

(3) Ninja: After installation, record the path and

http://next.matrix.jp/config-path-win7.html

Let's register the path of the environment variable by referring to

3. Download openpose

https://github.com/CMU-Perceptual-Computing-Lab/openpose/tree/windows Go to and press the green button on the top right- "Clone or download"-> "Download ZIP" to download openpose. Unzip the downloaded ZIP and move the entire second "openpose-windows" file to `C: /`

4. Install Caffe

Issue a command

`\openpose\3rdparty\caffe\caffe-Let's hit windows.`


 Then type `` `scripts \ build_win.cmd```.
 If there are no problems with the previous 1 to 3 steps, the installation will start successfully.
 And the installation time here is about 0.5 hours.

 And finally, the question:

#### **`Does D:\openpose\3rdparty\caffe\caffe-windows\build\..\..\include\caffe specify a file name or directory name on the target (F = file, D = directory)?,If you get, answer "D".`**

5. Launch sln in Visual Studio

`{openpose_path}\windows_project\OpenPose.sln.With the path of "OpenPose.Double-click "sln" to start up.`




 Next, select the one in the middle, set "Debug" to "Release", and compile.

## 6. Remaining files
(1)COCO model
 Go to the site with the blue letters "COCO model" and put "pose_iter_440000.caffemodel" in place of ``` {openpose_folder} \ models \ pose \ coco \` ``.

(2) MPI model (15 key-points, faster and less memory than COCO) 
 Go to the site in blue letters "MPI model (15 key-points, faster and less memory than COCO)" and go to the site.

#### **`{openpose_folder}\models\pose\mpi\In the place of "pose"_iter_160000.Let's put "caffe model".`**
```Let's put "caffe model".


### (3) Take out all the contents and copy and plague
```{openpose_folder}\3rdparty\caffe\caffe-windows\build\install\bin\```Copy everything in```{openpose_folder}\windows_project\x64\release```Let's copy to

### (4) Copy and plague 3 files

#### **`{openpose_folder}\3rdparty\caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\x64\vc14\bin\It is in`**

1.opencv_ffmpeg310_64.dll 2.opencv_video310.dll 3.opencv_videoio310.dll

Copy and

{openpose_folder}\windows_project\x64\releaseLet's plague.

Let's run

With a command

`\openpose-windows\windows_project\x64\Hit Release,`



#### **`.\OpenPoseDemo.Let's hit the exe.`**

And if there is no problem with the previous setting, you should be able to attend.

Try other settings (drop "Esc" when you're done)

** Be sure to type cd C: \ openpose-windows \ windows_project \ x64 \ Release and then execute the following command **

.\OpenPoseDemo.exe --net_resolution 320x176 Hitting will increase the FPS and reduce the delay (please pay attention to the specifications of your computer).

.\OpenPoseDemo.exe --video examples/media/video.avi When you hit, the image of the model is loaded and the analysis starts automatically.

.\OpenPoseDemo.exe --write_pose_json output/ When you hit, WebCam opens, and by the number of seconds, the location information of the person's skeleton

`\openpose-windows\windows_project\x64\Release\It is saved up to output.`



### If you hit "1-9" on the keyboard while it is running, only the numbers you have decided will be shown.


 I often use:

#### **`.\OpenPoseDemo.exe --write_pose_json output/ --image_dir examples\photo`**

[PYTHON] [OpenPose] installation method and compilation details (about 4 hours)