Jetson Nano

本文主要用于记录，基于Jetson Nano开发板上应用于无人驾驶小车的系统开发流程，主要包括镜像制作、环境配置、硬件驱动等。

1. 镜像制作

镜像下载

重点提醒：JetsonNano处理器架构是aarch64架构，所以在jetson nano上安装软件时请选择arrch64版本的，否则会导致无法预料的严重后果。
arm64和aarch64之间的区别：arm64已经与aarch64合并，因为aarch64和arm64指的是同一件事。ARM64是由Apple创建的，而AARCH64是由其他人（最著名的是GNU / GCC的）创建的。用于aarch64的Apple后端称为arm64，而LLVM 编译器社区开发的后端称为aarch64（因为它是64位ISA的规范名称），后来将arm64和 aarch64 两者合并，现在的后端称为aarch64。

镜像文件

Jetson 下载中心6

下载完成后本地会保存为后缀为.zip的文件

格式化SD卡

初始化一张大于32GB的SD卡，并对其进行格式化，格式可以为exFAT

镜像烧录

文章以Mac OS系统为例进行镜像的烧录，下载树莓派镜像烧录软件，软件安装完成后，操作系统选项里点击使用自定义镜像并选择已下载的镜像文件，SD卡选项选择要进行烧录的SD卡即可，设置完成后点击烧录按钮镜像的烧录，等待完成即可。

2. 环境配置

修改环境变量

Jetson Nano利用官方镜像进行安装后，系统已经安装好了JetPack，cuda，cudaa，OpenCV等组件，需要修改下环境变量才可以使用。

利用gedit打开 ~ 路径下.bashrc文件：
```
sudo gedit ~./bashrc
```
文件的最后添加以下三行：
```
export PATH=/usr/local/cuda-10.2/bin:$PATH
   
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
   
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda-10.2
```
重新执行.bashrc文件，直接生效；

source ~./bashrc

输入nvcc -V命令进行测试，如果显示如下信息，证明修改正确。
```
nvcc -V
   
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_21:14:42_PDT_2019
Cuda compilation tools, release 10.2, V10.2.89
```

更换国内镜像源

NVIDIA官方提供的Linux镜像版本为Ubuntu 18.04 LTS，镜像默认的是Ubuntu官方源，在国内使用该源下载程序时速度较慢，所以需要更换。

首先，打开/etc/apt/sources.list文件，注释原内容

sudo gedit /etc/apt/sources.list

在末尾添加下述内容（以清华大学镜像源为例，注意镜像源需要支持arm64架构）

# 默认注释了源码镜像以提高 apt update 速度，如有需要可自行取消注释
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic main restricted universe multiverse
# deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic main restricted universe multiverse
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-updates main restricted universe multiverse
# deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-updates main restricted universe multiverse
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-backports main restricted universe multiverse
# deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-backports main restricted universe multiverse
deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-security main restricted universe multiverse
# deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-security main restricted universe multiverse
   
# 预发布软件源，不建议启用
# deb https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-proposed main restricted universe multiverse
# deb-src https://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ bionic-proposed main restricted universe multiverse

保存并关闭文件，然后输入以下命令来更新源：

sudo apt-get update

其它部分常用镜像源：

中科大镜像源：

# 默认注释了源码镜像以提高 apt update 速度，如有需要可自行取消注释
deb http://mirrors.ustc.edu.cn/ubuntu-ports bionic main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports bionic main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu-ports bionic-updates main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports bionic-updates main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu-ports bionic-backports main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports bionic-backports main restricted universe multiverse
deb http://mirrors.ustc.edu.cn/ubuntu-ports bionic-security main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports bionic-security main restricted universe multiverse
   
# 预发布软件源，不建议启用
# deb http://mirrors.ustc.edu.cn/ubuntu-ports bionic-proposed main restricted universe multiverse
# deb-src http://mirrors.ustc.edu.cn/ubuntu-ports bionic-proposed main restricted universe multiverse

阿里源：

# 默认注释了源码镜像以提高 apt update 速度，如有需要可自行取消注释
deb https://mirrors.aliyun.com/ubuntu-ports/ bionic main restricted universe multiverse
# deb-src https://mirrors.aliyun.com/ubuntu-ports/ bionic main restricted universe multiverse
deb https://mirrors.aliyun.com/ubuntu-ports/ bionic-updates main restricted universe multiverse
# deb-src https://mirrors.aliyun.com/ubuntu-ports/ bionic-updates main restricted universe multiverse
deb https://mirrors.aliyun.com/ubuntu-ports/ bionic-backports main restricted universe multiverse
# deb-src https://mirrors.aliyun.com/ubuntu-ports/s bionic-backports main restricted universe multiverse
deb https://mirrors.aliyun.com/ubuntu-ports/ bionic-security main restricted universe multiverse
# deb-src https://mirrors.aliyun.com/ubuntu-ports/ bionic-security main restricted universe multiverse
   
# 预发布软件源，不建议启用
# deb https://mirrors.aliyun.com/ubuntu-ports/ bionic-proposed main restricted universe multiverse
# deb-src https://mirrors.aliyun.com/ubuntu-ports/ bionic-proposed main restricted universe multiverse

3. pip安装

sudo apt-get install python3-pip python3-dev

更新pip，保证是最新版本

pip3 install --upgrade pip

4. Jtop安装

依赖安装

sudo apt-get install git cmake
sudo apt-get install python3-dev
sudo apt-get install libhdf5-serial-dev hdf5-tools
sudo apt-get install libatlas-base-dev gfortran

安装jtop
```
sudo -H pip3 install -U jetson-stats
```

5. 安装Tensorflow

安装依赖库

sudo apt install python3-scipy -y
sudo apt install python3-pandas -y
sudo apt install python3-sklearn -y
sudo apt install python3-seaborn -y

安装一些其他依赖

sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev

安装numpy

需要先升级 protobuf 版本至 3.3.0：
```
sudo pip3 install protobuf==3.3.0
```
安装 numpy：
```
sudo pip3 install -U numpy==1.16.1
```
注意这里numpy的版本，太高或者太低多多少少都会导致一些后面安装的问题，如果不指定版本，默认安装1.19.5版本会导致后面装h5py的时候报非法指令的错误。
安装h5py

这个包对依赖的版本要求真的非常高！出了各种各样的问题，最终选择下载源文件手动安装

笔记本在清华的镜像软件源下载h5py==2.10.0（h5py-2.10.0.tar.gz）Links for h5py (tsinghua.edu.cn)

下载完成用U盘拷贝到nano中

在nano里面解压文件，并在对应文件夹中（文件夹可以看到setup.py）打开终端

注意！在安装前先要安装Cython==3.0.0a10!如果有Cython也要保证版本一致！
```
sudo pip3 install -U Cython==3.0.0a10
```
接着运行setup.py来装h5py，注意这里的python3不能少，否则系统默认是2.7的版本，会找不到Cython
```
sudo python3 ./setup.py install
```
一段漫长的编译以后，h5py==2.10.0就装好了！

安装keras-applications包

sudo pip3 install -U keras-applications

安装future包
```
sudo pip3 install -U future
```

安装 setuptools 和 testresources 包

sudo pip3 install -U setuptools testresources

检查相应库的版本

最后检查一遍相应库的版本看是否满足安装 Tensorflow 的条件。

检查相应库的版本：
```
pip3 list
```
安装 Tensorflow 的条件：
```
numpy==1.16.1, future==0.17.1, mock==3.0.5, h5py==2.9.0, gast==0.2.2, keras_preprocessing==1.0.5, keras_applications==1.0.8, scipy==1.4.1
```
安装Tensorflow

使用pip3安装TensorFlow，该命令将安装与JetPack 4.4兼容的TensorFlow的最新版本：
```
sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v461 tensorflow
```
如果速度太慢可以pip暂时换源：（在上面命令后加 -i https://pypi.tuna.tsinghua.edu.cn/simple）
```
sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v461 tensorflow -i https://pypi.tuna.tsinghua.edu.cn/simple
```
等待一段时间后，Tensorflow就装好啦！！

安装Torch

使用pip3安装torch与torchvision

sudo pip3 install torch==1.8.1
sudo pip3 install torchvision==0.9.1

6. 性能设置

除了无桌面模式和swap虚拟内存外，还可以提高功耗和解除硬件限制来提高性能。

设置功耗模式： 查询现有模式
```
sudo nvpmodel -q
```
设定高功率（10W）模式
```
sudo nvpmodel -m 0
```
设定低功率（5W）模式
```
sudo nvpmodel -m 1
```

改变模式之后，可以再执行第一个”查询“指令，看看是否改变

解除硬件限制：以前Jetson系列产品 ${home} 目录下，都会有个 jetson_clocks.sh 脚步，可以用来将性能压榨到最极致，而 Nano 的 {home} 目录找不到了。。。其实 jetson_clock 没有消失，只是藏起来了。您在任何地方，只要执行以下指令就可以解除硬件限制。
```
sudo jetson_clocks
```

7. OpenCV

安装比较麻烦，且容易失败，如无必要可继续使用原生opencv

卸载原生opencv

sudo apt-get purge libopencv*
sudo apt autoremove
sudo apt-get update
sudo apt-get install -y build-essential pkg-config

通过opencv本地包进行安装

* pip官网下载opencv-python=4.5.1.48
* 直接下载文件opencv_python-4.5.1.48-cp36-cp36m-manylinux2014_aarch64.whl
* 本地安装完成后，安装numpy==1.16.1

通过opencv项目编译安装（不推荐）

安装v4l2
```
sudo apt install v4l-utils 
```
```

https://github.com/yqlbu/jetson-packages-family#opencv

列表中找到opencv并根据提示进行安装
其中opencv，opencv_contrib，opencv_extra 可以先下载到home目录下，以避免因网络不稳而无法安装 ```

安装脚本：

#!/bin/bash
# License: MIT. See license file in root directory
# Copyright(c) JetsonHacks (2017-2019)
# Modifier(c) Kevin Yu (2020)
  
echo -e "
+-------------------------------------------------------+
@@@ Welcome to OpenCV4.1-Jetson Installation Toolkit @@@@
|							|
|	  Select your Jetson DevKit Model		|
|							|
|	  1. Jetson-Nano				|	
|	  2. Jetson-TX2					|
|	  3. Jetson-Xavier				|
|							|
+-------------------------------------------------------+
"
  
param (){
read -p '[BASH]  Put your Jetson model # here: ' NUM
if [ "$NUM" == '1' ] || [ "$NUM" == '2' ] || [ "$NUM" == '3' ]
then
	echo -e "[BASH]  You choose #$NUM."
	read -p '[BASH]  Do you wish to continue? (y/n) ' VAL
	#if y continue, else done
	if [ "$VAL" == "y" ]
	then
		echo -e "[BASH]  Installation will start in 3 seconds."
		sleep 3
 		# Jetson Nano
		if [ "$NUM" == '1' ]; then
			ARCH_BIN=5.3
 		# Jetson TX2
		elif [ "$NUM" == '2' ]; then
			ARCH_BIN=6.2
		 # Jetson AGX Xaviers
		elif [ "$NUM" == '3' ]; then
			ARCH_BIN=7.2
		fi
		start
	else
		echo "[BASH]  Installation ends..."
	fi
else
	echo "[BASH]  Please put the right # "
	param
fi
}
  
start (){
DOWNLOAD_OPENCV_EXTRAS=YES
WHEREAMI=$PWD
NUM_JOBS=$(nproc)
CLEANUP=true
PACKAGE_OPENCV="-D CPACK_BINARY_DEB=ON"
CMAKE_INSTALL_PREFIX=$INSTALL_DIR
  
echo "[BASH]  Installation start!"
sleep 2
  
# Print out the current configuration
echo "[BASH]  Build configuration: "
echo " NVIDIA Jetson Nano"
echo " OpenCV binaries will be installed in: $CMAKE_INSTALL_PREFIX"
echo " OpenCV Source will be installed in: $OPENCV_SOURCE_DIR"
if [ "$PACKAGE_OPENCV" = "" ] ; then
   echo " NOT Packaging OpenCV"
else
   echo " Packaging OpenCV"
fi
  
if [ $DOWNLOAD_OPENCV_EXTRAS == "YES" ] ; then
 echo "[BASH]  Also downloading opencv_extras"
fi
  
# Repository setup
sudo apt-add-repository universe
sudo apt-get update
# Download dependencies for the desired configuration
cd $WHEREAMI
echo "[BASH]  Install dependencies ..."
sleep 2
sudo apt-get install -y \
    build-essential \
    cmake \
    libavcodec-dev \
    libavformat-dev \
    libavutil-dev \
    libeigen3-dev \
    libglew-dev \
    libgtk2.0-dev \
    libgtk-3-dev \
    libjpeg-dev \
    libpng-dev \
    libpostproc-dev \
    libswscale-dev \
    libtbb-dev \
    libtiff5-dev \
    libv4l-dev \
    libxvidcore-dev \
    libx264-dev \
    qt5-default \
    zlib1g-dev \
    libwebp-dev \
    pkg-config
  
# We will be supporting OpenGL, we need a little magic to help
# https://devtalk.nvidia.com/default/topic/1007290/jetson-tx2/building-opencv-with-opengl-support-/post/5141945/#5141945
cd /usr/local/cuda/include
sudo patch -N cuda_gl_interop.h $WHEREAMI'/patches/OpenGLHeader.patch' 
  
# Python 2.7
sudo apt-get install -y python-dev  python-numpy  python-py  python-pytest
# Python 3.6
sudo apt-get install -y python3-dev python3-numpy python3-py python3-pytest
  
# GStreamer support
sudo apt-get install -y libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev 
  
echo "[BASH]  Download sources ..."
sleep 2
  
cd $OPENCV_SOURCE_DIR
# git clone --branch "$OPENCV_VERSION" https://github.com/opencv/opencv.git
# git clone --branch "$OPENCV_VERSION" https://github.com/opencv/opencv_contrib.git
  
if [ $DOWNLOAD_OPENCV_EXTRAS == "YES" ] ; then
 echo "[BASH]  Installing opencv_extras"
 # This is for the test data
 cd $OPENCV_SOURCE_DIR
 # git clone https://github.com/opencv/opencv_extra.git
 cd opencv_extra
 git checkout -b v${OPENCV_VERSION} ${OPENCV_VERSION}
fi
  
# Patch the Eigen library issue ...
cd $OPENCV_SOURCE_DIR/opencv
sed -i 's/include <Eigen\/Core>/include <eigen3\/Eigen\/Core>/g' modules/core/include/opencv2/core/private.hpp
  
# Create the build directory and start cmake
cd $OPENCV_SOURCE_DIR/opencv
mkdir build
cd build
  
echo "[BASH]  Build from sources ..."
sleep 2
  
echo $PWD
time cmake -D CMAKE_BUILD_TYPE=RELEASE \
      -D CMAKE_INSTALL_PREFIX=${CMAKE_INSTALL_PREFIX} \
      -D WITH_CUDA=ON \
      -D CUDA_ARCH_BIN=${ARCH_BIN} \
      -D CUDA_ARCH_PTX="" \
      -D ENABLE_FAST_MATH=ON \
      -D CUDA_FAST_MATH=ON \
      -D WITH_CUBLAS=ON \
      -D WITH_LIBV4L=ON \
      -D WITH_V4L=ON \
      -D WITH_GSTREAMER=ON \
      -D WITH_GSTREAMER_0_10=OFF \
      -D OPENCV_DNN_CUDA=ON \
      -D WITH_QT=ON \
      -D WITH_TBB=ON \
      -D BUILD_opencv_python2=ON \
      -D BUILD_opencv_python3=ON \
      -D BUILD_TESTS=OFF \
      -D BUILD_PERF_TESTS=OFF \
      -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
      $"PACKAGE_OPENCV" \
      ../
  
  
if [ $? -eq 0 ] ; then
  echo "[BASH]  CMake configuration make successful"
else
  # Try to make again
  echo "[BASH]  CMake issues " >&2
  echo "[BASH]  Please check the configuration being used"
  exit 1
fi
  
# Consider the MAXN performance mode if using a barrel jack on the Nano
time make -j$NUM_JOBS
if [ $? -eq 0 ] ; then
  echo "[BASH]  OpenCV make successful"
else
  # Try to make again; Sometimes there are issues with the build
  # because of lack of resources or concurrency issues
  echo "[BASH]  Make did not build " >&2
  echo "[BASH]  Retrying ... "
  # Single thread this time
  make
  if [ $? -eq 0 ] ; then
    echo "[BASH]  OpenCV make successful"
  else
    # Try to make again
    echo "[BASH]  Make did not successfully build" >&2
    echo "[BASH]  Please fix issues and retry build"
    exit 1
  fi
fi
  
echo "[BASH]  Installing ... "
sudo make install
sudo ldconfig
if [ $? -eq 0 ] ; then
   echo "[BASH]  OpenCV installed in: $CMAKE_INSTALL_PREFIX"
else
   echo "[BASH]  There was an issue with the final installation"
   exit 1
fi
  
# If PACKAGE_OPENCV is on, pack 'er up and get ready to go!
# We should still be in the build directory ...
if [ "$PACKAGE_OPENCV" != "" ] ; then
   echo "[BASH]  Starting Packaging"
   sudo ldconfig  
   time sudo make package -j$NUM_JOBS
   if [ $? -eq 0 ] ; then
     echo "[BASH]  OpenCV make package successful"
   else
     # Try to make again; Sometimes there are issues with the build
     # because of lack of resources or concurrency issues
     echo "[BASH]  Make package did not build " >&2
     echo "[BASH]  Retrying ... "
     # Single thread this time
     sudo make package
     if [ $? -eq 0 ] ; then
       echo "[BASH]  OpenCV make package successful"
     else
       # Try to make again
       echo "[BASH]  Make package did not successfully build" >&2
       echo "[BASH]  Please fix issues and retry build"
       exit 1
     fi
   fi
fi
  
# Install Python packages
sudo apt-get install -y libopencv-dev
sudo apt-get install -y libopencv-core-dev
  
# check installation
IMPORT_CHECK="$(python3 -c "import cv2 ; print(cv2.__version__)")"
if [[ $IMPORT_CHECK != *$OPENCV_VERSION* ]]; then
  echo "[BASH]  There was an error loading OpenCV in the Python sanity test."
  echo "[BASH]  The loaded version does not match the version built here."
  echo "[BASH]  Please check the installation."
  echo "[BASH]  The first check should be the PYTHONPATH environment variable."
fi
  
echo "[BASH]    Installation completed ..."
sleep 2
  
}
  
# Execute
ARCH_BIN=0
INSTALL_DIR=/usr/local
OPENCV_SOURCE_DIR=$HOME
OPENCV_VERSION=4.1.1
param
  
  
  

参考：https://blog.csdn.net/qq_68080661/article/details/132057624

常见问题

Illegal instruction (core dumped)
- 终端运行
```
sudo vim ~/.bashrc
```
- 末尾添加环境变量，保存并关闭文件
```
export OPENBLAS_CORETYPE=ARMV8
```
- 终端运行
```
source ~/.bashrc
```

Cannot identify device ‘/dev/video…

* pip官网下载opencv-python=4.5.1.48
* 直接下载文件opencv_python-4.5.1.48-cp36-cp36m-manylinux2014_aarch64.whl
* 本地安装完成后，安装numpy==1.16.1

https://blog.csdn.net/qq_56548850/article/details/123967404

https://blog.csdn.net/weixin_45235219/article/details/128840418

Kuture

Jetson Nano

Jetson Nano

1. 镜像制作

2. 环境配置

3. pip安装

4. Jtop安装

5. 安装Tensorflow

6. 性能设置

7. OpenCV

常见问题

Search

Table of Contents