Manifold-2G tensorflow,gpu环境配置

Manifold-2G tensorflow,gpu环境配置

1.Jetpack3.3 Opencv3.3.1 python3.5和python2.7(python执行python2.7，python3执行python3.5 pip对应python2.7 pip3 对应python3.5)
2.tensorflow_gpu-1.9.0+nv18.8-cp35-cp35m-linux_aarch64(cp35 代表python3.5)

官方的版本：
https://developer.nvidia.com/embedded/downloads#?tx=$product,jetson_tx2

Jetpack3.3只支持tensorflow1.9和tensorflow1.14;
Tensorflow1.14需要cuda10，但是jetpack3.3不支持cuda10，升级jetpack需要Ubuntu升级18.04，但是ROS kinetic 只支持Ubuntu16.04,综上所述：
Maniflod2G的版本为:Jetpack3.3,tensorflow-gpu-1.9,python3.5,
opencv3.3.1.

二、Opencv3.3.1安装
[注意]执行命令前可以先查看下自身环境版本。(目的：产生C++版本和python版本的opencv库，arm架构下没有现成的opencv-python包)整个安装时间大约2个小时。
[1]sudo apt-get install --only-upgrade g+±5 cpp-5 gcc-5
[2]基于jetson安装opencv依赖项
sudo apt-get install build-essential make cmake cmake-curses-gui
g++ libavformat-dev libavutil-dev
libswscale-dev libv4l-dev libeigen3-dev
libglew-dev libgtk2.0-dev

[3] gstreamer安装依赖项
sudo apt-get install libdc1394-22-dev libxine2-dev
libgstreamer1.0-dev
libgstreamer-plugins-base1.0-dev
[4]安装其他依赖项
sudo apt-get install libjpeg8-dev libjpeg-turbo8-dev libtiff5-dev
libjasper-dev libpng12-dev libavcodec-dev
sudo apt-get install libxvidcore-dev libx264-dev libgtk-3-dev
libatlas-base-dev gfortran
sudo apt-get install libopenblas-dev liblapack-dev liblapacke-dev
[5]开始安装
5.1使用dji用户登录系统
5.2解压third-party-library-folder压缩包
5.3 cd 到opencv目录下，创建build目录
mkdir build
5.4 cd 到build 目录
执行：
cmake -D CMAKE_BUILD_TYPE=RELEASE
-D ENABLE_CXX11=ON
-D CMAKE_INSTALL_PREFIX=/usr/local
-D INSTALL_C_EXAMPLES=ON
-D INSTALL_PYTHON_EXAMPLES=ON
-D WITH_TBB=ON
-D WITH_V4L=ON
-D WITH_QT=ON
-D WITH_OPENGL=ON
-D WITH_CUDA=ON
-D WITH_VTK=ON
-D WITH_OPENGL=ON
-D OPENCV_EXTRA_MODULES_PATH=…/…/opencv_contrib/modules
-D BUILD_EXAMPLES=ON …
5.5 make -j4 然后 sudo make install
5.6 安装完成，可以查看opencv版本
pkg-config --modversion opencv

编译时候可能产生的错误：

三、tensorflow1.9-gpu
1、安装步骤
[1]必须创建新用户如dev，登录系统
adduser dev
ros环境默认py2.7，opencv环境需要py3.5
[2]cd到tensorflow_gpu-1.9.0+nv18.8-cp35-cp35m-linux_aarch64.whl目录
[3]pip3 install tensorflow_gpu-1.9.0+nv18.8-cp35-cp35m-linux_aarch64.whl
[4]安装完成，测试运行
import tensorflow as tf
hello=tf.constant(‘hello’)
sess=tf.Session()
print (sess.run(hello))
输出”hello”且未报错即安装成功。

六、编译大疆SDK注意事项
OPENCV 报cuda不存在时
报错原因：gedit ~/catkin_ws/src/Onboard-SDK-ROS/dji_sdk_demo/CMakeLists.txt
if (OPEN_CV_3_3_0_INSTALLED AND FOUND_OPENCV_VIZ AND darknet_ros_FOUND AND FOUND_ADVANCED_SENSING) message(STATUS “Found OpenCV ${OpenCV_VERSION}, Viz3d, darknet_ros, and advanced sensing module, depth percpetion demo will be compiled”) set(OpenCV_INCLUDE_DIRS /usr/local/include /usr/local/include/opencv2 ) set(OpenCV_LIB_DIR /usr/local/lib ) add_executable(demo_stereo_object_depth_perception src/demo_stereo_object_depth_perception.cpp ${STEREO_UTILITY_SRC_FILE} ) target_link_libraries(demo_stereo_object_depth_perception ${catkin_LIBRARIES} ${DJIOSDK_LIBRARIES} ${OpenCV_LIBS} ) add_dependencies(demo_stereo_object_depth_perception dji_sdk_generate_messages_cpp) endif()

解决办法：在CMakeLists.txt 中添加（添加上面编译过的opencv）
cmake_minimum_required(VERSION 2.8.3)
set(OpenCV_DIR “/home/dji/software/third-party-library-folder/opencv/build”) project(dji_sdk_demo)