Tx2上人体姿态估计AlphaPose配置安装教程

一、Tx2的刷机
电脑主机（Linux系统）
实现方法有两种：
①安装Linux系统
②下载VMware安装虚拟机（个人觉得用虚拟机较为方便）
刷机过程：
主机在官网https://developer.nvidia.com/embedded/jetpack下载JetPack安装包，例如JetPack-L4T-3.3-linux-x64.run，给文件加可执行权限并运行：

sudo chmod +x JetPack-L4T-3.3-linux-x64.run
./JetPack-L4T-3.3-linux-x64.run

然后主机就开始下载相应的安装包，这个过程可能需要30min左右，如果出现error的情况可以检测一下网络连接是否正常
下载完成之后进入如图所示的界面




执行下一步后，会出现一个提示，按照步骤操作，使Tx2进入recovery mode，翻译成中文如下：
（1）Tx2断电
（2）用Micro USB数据线连接PC和Tx2。（注意：数据线接到Tx2的Micro USB接口J20）
（3）接通Tx2电源
（4）按下并松开Tx2的POWER键S4；按下并保持FORCE RECOVERY键S3；在此期间按下并松开RESET键S1；等待2秒后松开FORCE RECOVERY键S3。
可以在终端输入lsusb，如果发现Nvidia Corp则说明Tx2与主机已连接，按照提示按下Enter键等待安装即可，之后终端界面运行到查找Tx2的IP的地方会停止，此时关闭该窗口（此处不要删除下载的文件），再次按照之前的操作打开安装界面，将Flash OS image to target 改成no action，点击next，会出现输入Tx2的IP地址的界面，由于已经安装好了图形界面，此时可以将Tx2接显示屏。然后将电脑主机和Tx2通过网线连接至同一个路由器上
PS:如果没有路由器的话可以尝试利用WiFi或者手机开热点的方式将主机和Tx2连接至同一个网络，然后通过Tx2的network选项查找IP地址并将IP地址，login（nvidia），password（nvidia）输入点击next即可，如果出现错误可以根据提示检查一下IP地址是否输入正确或者重试。
之后就是漫长的安装过程啦！

二、安装AlphaPose
根据Tx2自带的python2.7和python3.5主要有两种版本（General version和Pytorch version)，这里主要分享搭建Pytorch version
根据官方文档https://github.com/MVIG-SJTU/AlphaPose/tree/pytorch

git clone -b pytorch https://github.com/MVIG-SJTU/AlphaPose.git
cd AlphaPose
python3 demo.py --indir ${img_directory}$ --outdir examples/res 

报错：no module named torch说明没有安装pytorch

sudo apt install libopenblas-dev libatlas-dev liblapack-dev
sudo apt install liblapacke-dev checkinstall # For OpenCV
sudo apt-get install python3-pip

pip3 install --upgrade pip3==9.0.1
sudo apt-get install python3-dev

sudo pip3 install numpy scipy # 这个花费的时间略长，20分钟到30分钟左右
sudo pip3 install pyyaml
sudo pip3 install scikit-build
sudo apt-get -y install cmake
sudo apt install libffi-dev
sudo pip3 install cffi

sudo gedit ~/.bashrc
export CUDNN_LIB_DIR=/usr/lib/aarch64-linux-gnu
export CUDNN_INCLUDE_DIR=/usr/include
source ~/.bashrc

git clone http://github.com/pytorch/pytorch
cd pytorch
sudo pip3 install -U setuptools
sudo pip3 install -r requirements.txt
git submodule update --init --recursive

sudo nvpmodel -m 0         # 切换工作模式到最大
sudo  ~/jetson_clocks.sh   # 强制开启风扇最大转速

python3 setup.py build_deps
NO_SYSTEM_NCCL=1 DEBUG=1 sudo python3 setup.py build develop

sudo apt clean
sudo apt-get install libjpeg-dev zlib1g-dev

git clone https://github.com/python-pillow/Pillow.git
cd Pillow/
sudo python3 setup.py install
sudo pip3 install pandas # ~20-30 min
sudo pip3 install Cython
sudo pip3 install scikit-image

sudo pip3 --no-cache-dir install torchvision

参考https://m.oldpan.me/archives/nvidia-jetson-tx2-source-build-pytorch
此时安装的pytorch版本为pytorch v1.0，但是官方文档要求pytorch==0.4.0
参考https://github.com/lixiandea/AlphaposeWithPytorch1.0
报错：no module named cv2说明没有安装opencv

cd
sudo apt-get install -y \
    libglew-dev \
    libtiff5-dev \
    zlib1g-dev \
    libjpeg-dev \
    libpng12-dev \
    libjasper-dev \
    libavcodec-dev \
    libavformat-dev \
    libavutil-dev \
    libpostproc-dev \
    libswscale-dev \
    libeigen3-dev \
    libtbb-dev \
    libgtk2.0-dev \
    cmake \
    pkg-config
# Python 2.7
sudo apt-get install -y python-dev python-numpy python-py python-pytest -y
#GStreamer support
sudo apt-get install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev 
git clone https://github.com/opencv/opencv.git
cd opencv
git checkout -b v3.3.0 3.3.0
# This is for the test data
cd $HOME
git clone https://github.com/opencv/opencv_extra.git
cd opencv_extra
git checkout -b v3.3.0 3.3.0
cd opencv
mkdir build
cd build
# Jetson TX2 
cmake \
    -DCMAKE_BUILD_TYPE=Release \
    -DCMAKE_INSTALL_PREFIX=/usr \
    -DBUILD_PNG=OFF \
    -DBUILD_TIFF=OFF \
    -DBUILD_TBB=OFF \
    -DBUILD_JPEG=OFF \
    -DBUILD_JASPER=OFF \
    -DBUILD_ZLIB=OFF \
    -DBUILD_EXAMPLES=ON \
    -DBUILD_opencv_java=OFF \
    -DBUILD_opencv_python2=ON \
    -DBUILD_opencv_python3=ON \
    -DENABLE_PRECOMPILED_HEADERS=OFF \
    -DWITH_OPENCL=OFF \
    -DWITH_OPENMP=OFF \
    -DWITH_FFMPEG=ON \
    -DWITH_GSTREAMER=ON \
    -DWITH_GSTREAMER_0_10=OFF \
    -DWITH_CUDA=ON \
    -DWITH_GTK=ON \
    -DWITH_VTK=OFF \
    -DWITH_TBB=ON \
    -DWITH_1394=OFF \
    -DWITH_OPENEXR=OFF \
    -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-8.0 \
    -DCUDA_ARCH_BIN=6.2 \
    -DCUDA_ARCH_PTX="" \
    -DINSTALL_C_EXAMPLES=ON \
    -DINSTALL_TESTS=ON \
    -DOPENCV_TEST_DATA_PATH=../opencv_extra/testdata \
    ../
# Consider using all 6 cores; $ sudo nvpmodel -m 2 or $ sudo nvpmodel -m 0
make -j4
-DBUILD_opencv_python2=ON \
-DBUILD_opencv_python3=ON \
#把opencv的python2,python3的接口都打开
chmod +x buildOpenCV.sh
./buildOpenCV.sh
sudo make install

参考https://blog.csdn.net/w5688414/article/details/78771958
报错：no module named matplotlib
sudo pip3 install matplotlib
报错：no module named pyzmq

git clonehttps://github.com/zeromq/libzmq
./autogen.sh
sudo apt-get install libtool autoconf automake
./configure
make -j 4
make check
make install
sudo ldconfig
sudo pip3 install pyzmq

参考https://blog.csdn.net/n_sev7/article/details/77320250
报错：FileNotFoundError: [Errno 2] No such file or directory: ‘models/yolo/yolov3-spp.weights’
把官网上的model下载放到指定文件夹
https://github.com/MVIG-SJTU/AlphaPose/tree/pytorch
报错：RuntimeError: cuda runtime error (71) : operation not supported at /home/nvidia/pytorch/torch/csrc/generic/StorageSharing.cpp:232

python3 demo.py  --indir ${img_directory} --outdir examples/res --sp#取消多线程

同样如果在运行video_demo.py是出现cv2 can’t pickle 问题时也可以在末尾加上–sp取消多线程
报错：ImportError: /usr/lib/python3.5/dist-packages/cv2.cpython-35m-aarch64-linux-gnu.so: undefined symbol: _ZTIN2cv3dnn19experimental_dnn_v15LayerE

cd opencv/build
sudo make install

最后衷心的感谢以上给予帮助的人！！

---

参考资料
https://github.com/MVIG-SJTU/AlphaPose/tree/pytorch
https://blog.csdn.net/n_sev7/article/details/77320250
https://blog.csdn.net/w5688414/article/details/78771958
https://github.com/lixiandea/AlphaposeWithPytorch1.0
https://m.oldpan.me/archives/nvidia-jetson-tx2-source-build-pytorch
https://github.com/kleinzcy/install-note/blob/master/TX2.md