交叉编译ARM平台、移植ArmNN

参考文章：

检查lib文件的软连接方向：

ls -l libboost_system.so

进入板子linux系统：

ssh [email protected]
ping 10.6.5.134

用ssh将本地资源复制到服务器

https://blog.csdn.net/qq_39377418/article/details/89367041

//查看当前系统磁盘使用空间

df -h

//查看当前目录文件占用空间大小

du -sh *

Linux -- 查询某个文件夹下的文件数量

https://www.cnblogs.com/gengyufei/p/12804376.html

ubuntu16.04系统安装且挂载硬盘且设为开机启动项

https://blog.csdn.net/weixin_42652125/article/details/81171233

命令行批量替换字符串

sed命令替换字符包含斜杠\，引号的处理方法 - franjia - 博客园

https://www.cnblogs.com/franjia/p/6690060.html

【转载】linux中批量替换文本中字符串命令 - -天道酬勤- - 博客园

https://www.cnblogs.com/wbl001/p/12492055.html

sed -i 's#home/gavin/桌面/TOF_Doc/libroyale_ningbosunnyopotech#home/robert/DeepLearning/royale#g' `grep home/gavin/桌面/TOF_Doc/libroyale_ningbosunnyopotech -rl *`

杀死从文件夹启动的可执行程序

查序号，https://www.cnblogs.com/hml-blog-com/p/11558369.html

ps aux
kill -s 9 3960

Linux创建连接命令 ln -s创建软连接

https://www.cnblogs.com/zhangna1998517/p/11347364.html

ARM交叉编译工具链分类说明

https://blog.csdn.net/qq_16149777/article/details/82349868

命令行分区

linux将其他分区空间划分给root分区的方法（ext4文件系统）

https://blog.csdn.net/qq_26514685/article/details/106673825

linux如何查看磁盘剩余空间

https://www.cnblogs.com/jasonxu19900827/p/5282237.html

在ARM Linux下挂载SD卡分区

https://blog.csdn.net/timeless_2014/article/details/82319554

ubuntu 硬盘管理工具

https://blog.csdn.net/bluebird_shao/article/details/8241698

ubuntu16.04使用GParted对/根目录扩容

https://blog.csdn.net/Carina_Cao/article/details/90270389

linux 如何查看硬盘大小,存储空间大小等系统信息及硬件信息

https://www.cnblogs.com/lixuejian/p/12100166.html

Linux下tempfs及/dev/shm原理与应用

https://blog.csdn.net/guo8113/article/details/28590963

烧SD卡和通信

Ubuntu下安装ssh与配置

https://www.cnblogs.com/cookiewu/p/9664062.html

如何在Ubuntu下向TF/SD卡中烧写镜像（程序），较好

https://blog.csdn.net/zyc_csdn/article/details/89062686

Ubuntu14.04下tftp安装，并与开发板通信

https://blog.csdn.net/smz958150/article/details/75307463

ubuntu下的tftp上传和下载操作方法

https://blog.csdn.net/long0814long0814/article/details/65477673

ubantu下ssh: connect to host xxxxxxxxxx port 22: Connection refused问题

https://blog.csdn.net/yundou520/article/details/82658809

如何在ubuntu中通过串口访问开发板（如：树莓派），putty

https://blog.csdn.net/dgj8300/article/details/51045874/

ubuntu下串口工具的安装与使用，minicom

https://blog.csdn.net/lzhitwh/article/details/80304579

深度学习与程序移植

深度强化学习实验中的paper绘图方法

https://mp.weixin.qq.com/s/94nwkxn20ZixmcsX8oUJIw

dair-ai/ml-visuals: Visuals contains figures and templates which you can reuse and customize to improve your scientific writing.

https://github.com/dair-ai/ml-visuals

Linux 中最常用 150 个命令汇总

https://mp.weixin.qq.com/s/Lbp6_pz_UYrrGUtdQKwfBQ

Linux指定动态库搜索路径五种方法及优先级

https://blog.csdn.net/wwwtovvv/article/details/41477249

linux下cmake-3.11.4安装

https://blog.csdn.net/Jeffxu_lib/article/details/80712414

小技巧

linux查看可执行程序的链接库及文件位置

https://blog.csdn.net/weixin_42404296/article/details/82632354

Linux下使用SSH进行远程登录主机及操作

https://blog.csdn.net/Hellowenpan/article/details/82904109

linux下如何查看cpu信息

https://blog.csdn.net/liuli9/article/details/84112680

shell脚本中的条件测试if中的-z到-d的意思

https://blog.csdn.net/xiaodingqq/article/details/79981372

移植boost_1_55_0至arm的方法小结（有错误）

https://blog.csdn.net/dotphoenix/article/details/41114523

arm交叉编译器编译boost库并调用（有错误）

https://blog.csdn.net/jiangheng0535/article/details/18225775

移植ArmNN

在imx8qm上开心把玩Machine Learning（写得较简单，有一点参考意义）

https://blog.csdn.net/u013739490/article/details/84314290

Configuring the Arm NN SDK build environment for Caffe（较完善，主要还是参考官方编译教程）

https://blog.csdn.net/liugan528/article/details/80272763

ARM NN:Ubuntu 14.04 Caffe和TensorFlow以及TF-lite的ARM NN SDK编译环境搭建及MNIST程序测试（有一定参考意义）

https://blog.csdn.net/coinv2014/article/details/83582747

为Tensorflow、Tensorflow lite配置Arm NN SDK编译环境（有一定参考意义）

https://blog.csdn.net/a845414332/article/details/102718813

官方教程避坑：编译ARM NN/Tensorflow Lite

https://www.cnblogs.com/pepetang/p/10627542.html

在ARM板子上把玩Tensorflow Lite（操作复杂，暂无参考意义）

https://blog.csdn.net/computerme/article/details/80345065

ARMNN使用例程（有参考性，但不够完善）

https://github.com/ARM-software/ML-examples

官方指导网站

https://developer.arm.com/solutions/machine-learning-on-arm/developer-material/how-to-guides

我的编译代码，主要还是参考官方的编译文档：BuildGuideCrossCompilation.md

根据上述参考安装scons-2.4.1

protobuf version 3.5.2

build_x86_64.sh

#git submodule update --init --recursive
#./autogen.sh
mkdir x86_64_build
cd x86_64_build
../configure --prefix=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/x86_64_pb_install
make install -j16
cd ..

build_arm64.sh

mkdir arm64_build
cd arm64_build
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
../configure --host=aarch64-poky-linux \
    CC=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-gcc \
    CFLAGS="--sysroot=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux" \
    CXX=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-g++ \
    CXXFLAGS="--sysroot=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux" \
    --prefix=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install \
    --with-protoc=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/x86_64_pb_install/bin/protoc
make install -j16
cd ..

Build Boost library for arm64

build.sh

echo "building boost 1.64.0"
echo "using gcc : arm : aarch64-poky-linux-g++ ;" > user_config.jam
sh ./bootstrap.sh --prefix=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/boost_arm64_install
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
./b2 toolset=gcc-arm link=static cxxflags="-fPIC --sysroot=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux" --with-filesystem --with-test --with-log --with-program_options --user-config=user_config.jam
#./b2 install

Build Caffe for x86_64

参考：https://blog.csdn.net/yiyayiya557/article/details/105796768

和官方文档

Build Compute Library

build.sh

sudo dd if=/dev/zero of=/home/swap bs=64M count=16
sudo mkswap /home/swap
sudo swapon /home/swap
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
scons extra_cxx_flags="-fPIC --sysroot=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux" arch=arm64-v8a neon=1 opencl=1 embed_kernels=1 -j8 internal_only=0 # install_dir="/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/ComputeLibrary"
sudo swapoff /home/swap
sudo rm /home/swap

Build Tensorflow 1.15.3

build.sh

../armnn-branches-armnn_20_05/scripts/generate_tensorflow_protobuf.sh ../armnn-devenv/tensorflow-protobuf/ ../armnn-devenv/google/x86_64_pb_install

Build Flatbuffer 1.10.0

build_x86_64.sh

mkdir build_x86_64
cd build_x86_64
cmake .. -DFLATBUFFERS_BUILD_FLATC=1 \
	-DCMAKE_INSTALL_PREFIX=/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/x86_64_fb_install \
	-DFLATBUFFERS_BUILD_TESTS=0
make all install

build_arm64.sh

mkdir build_arm64
cd build_arm64
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
CC=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-gcc \
CXX=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-g++ \
cmake .. -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release -DCMAKE_TOOLCHAIN_FILE=../toolchain.cmake -DCMAKE_INSTALL_PREFIX=/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install
make all install

toolchain.cmake

# this is required
SET(CMAKE_SYSTEM_NAME Linux)

# specify the cross compiler
SET(CMAKE_C_COMPILER   /opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-gcc)
SET(CMAKE_CXX_COMPILER /opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-g++)

SET(CMAKE_SYSROOT /opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux)

# where is the target environment 
SET(CMAKE_FIND_ROOT_PATH  /opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux)

# specify the compiler flag
SET(CMAKE_C_FLAGS  -O3 -mtune=cortex-a35 -mcpu=cortex-a35 -march=armv8-a+fp+simd+crc+crypto)
SET(CMAKE_CXX_FLAGS  -std=c++11 -O3 -mtune=cortex-a35 -mcpu=cortex-a35 -march=armv8-a+fp+simd+crc+crypto)

# search for programs in the build host directories (not necessary)
SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
# for libraries and headers in the target directories
SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)

Build Onnx

build.sh

export LD_LIBRARY_PATH=/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/x86_64_pb_install/lib:$LD_LIBRARY_PATH
../armnn-devenv/google/x86_64_pb_install/bin/protoc ./onnx/onnx.proto --proto_path=. --proto_path=../armnn-devenv/google/x86_64_pb_install/include --cpp_out ../onnx

Build TfLite

build.sh

../flatbuffers-1.10.0/build_x86_64/flatc -c --gen-object-api --reflect-types --reflect-names schema.fbs

Build ArmNN

build.sh

mkdir build
cd build
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
cmake .. \
    -DARMCOMPUTE_ROOT=$HOME/DeepLearning/NXP-imx8QX/CompileResult/ArmComputeLibrary-master-v20.05 \
    -DARMCOMPUTE_BUILD_DIR=$HOME/DeepLearning/NXP-imx8QX/CompileResult/ArmComputeLibrary-master-v20.05/build/ \
    -DBOOST_ROOT=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/boost_arm64_install/ \
    -DARMCOMPUTENEON=1 -DARMCOMPUTECL=1 -DARMNNREF=1 \
    -DCAFFE_GENERATED_SOURCES=$HOME/DeepLearning/NXP-imx8QX/CompileResult/caffe-master-cpu/build/src \
    -DBUILD_CAFFE_PARSER=1 \
    -DONNX_GENERATED_SOURCES=$HOME/DeepLearning/NXP-imx8QX/CompileResult/onnx \
    -DBUILD_ONNX_PARSER=1 \
    -DTF_GENERATED_SOURCES=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/tensorflow-protobuf \
    -DBUILD_TF_PARSER=1 \
    -DBUILD_TF_LITE_PARSER=1 \
    -DTF_LITE_GENERATED_PATH=$HOME/DeepLearning/NXP-imx8QX/CompileResult/tflite \
    -DTF_LITE_SCHEMA_INCLUDE_PATH=$HOME/DeepLearning/NXP-imx8QX/CompileResult/tensorflow-1.15.3/tensorflow/lite/schema \
    -DFLATBUFFERS_ROOT=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install \
    -DFLATC_DIR=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install/bin \
    -DFLATBUFFERS_LIBRARY=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install/lib/libflatbuffers.a \
    -DFLATBUFFERS_INCLUDE_PATH=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install/include \
    -DPROTOBUF_ROOT=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install \
    -DPROTOBUF_INCLUDE_DIRS=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/include \
    -DPROTOBUF_LIBRARY_DEBUG=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/lib/libprotobuf.so.15.0.1 \
    -DPROTOBUF_LIBRARY_RELEASE=$HOME/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/lib/libprotobuf.so.15.0.1 \
    -DCMAKE_TOOLCHAIN_FILE=../toolchain.cmake \
    -DCMAKE_INSTALL_PREFIX=/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install
    #--no-warn-unused-cli
    #-DBUILD_TESTS=1

make -j$(nproc) install

toolchain.cmake

# this is required
SET(CMAKE_SYSTEM_NAME Linux)

# specify the cross compiler
SET(CMAKE_C_COMPILER   /opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-gcc)
SET(CMAKE_CXX_COMPILER /opt/fsl-imx-xwayland/4.14-sumo/sysroots/x86_64-pokysdk-linux/usr/bin/aarch64-poky-linux/aarch64-poky-linux-g++)

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall -Werror -Wold-style-cast -Wno-missing-braces -Wconversion -Wsign-conversion -pthread")

set(Boost_NO_BOOST_CMAKE ON)
set(Boost_USE_STATIC_LIBS ON)

SET(CMAKE_SYSROOT /opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux)

# where is the target environment 
SET(CMAKE_FIND_ROOT_PATH  /opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux)

# specify the compiler flag
SET(CMAKE_C_FLAGS  -O3 -mtune=cortex-a35 -mcpu=cortex-a35 -march=armv8-a+fp+simd+crc+crypto)
SET(CMAKE_CXX_FLAGS  -std=c++11 -O3 -mtune=cortex-a35 -mcpu=cortex-a35 -march=armv8-a+fp+simd+crc+crypto)

# search for programs in the build host directories (not necessary)
SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
# for libraries and headers in the target directories
SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)
SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)

测试armnn-mnist

build.sh

mkdir build
cd build
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux
aarch64-poky-linux-g++ -O3 --sysroot=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux -std=c++14 \
    -I/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install/include \
    ../mnist_caffe.cpp -o mnist_caffe \
    -L/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install/lib \
    -L/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/lib \
    -larmnn -larmnnCaffeParser -lprotobuf -lprotoc -lpthread

Makefile

ARMNN_LIB = ${HOME}/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install/lib
ARMNN_INC = ${HOME}/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install/include
PROTOBUF = ${HOME}/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/lib

all: mnist_caffe mnist_tf

mnist_caffe: mnist_caffe.cpp mnist_loader.hpp
	aarch64-poky-linux-g++ -O3 -std=c++14 -I$(ARMNN_INC) mnist_caffe.cpp -o mnist_caffe -L$(ARMNN_LIB) -L$(PROTOBUF) -lprotobuf -larmnn -larmnnCaffeParser -lpthread

mnist_tf: mnist_tf.cpp mnist_loader.hpp
	aarch64-poky-linux-g++ -O3 -std=c++14 -I$(ARMNN_INC) mnist_tf.cpp -o mnist_tf -L$(ARMNN_LIB) -L$(PROTOBUF) -lprotobuf -larmnn -larmnnTfParser -lpthread

clean:
	-rm -f mnist_tf mnist_caffe

test: mnist_caffe mnist_tf
	LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:$(ARMNN_LIB) ./mnist_caffe
	LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:$(ARMNN_LIB) ./mnist_tf

测试armnn-mobilenet-quant

build.sh

#!/bin/bash
set -e # Exit immediately if a command exits with a non-zero status.
BuildDir=build
if [ ! -d "$BuildDir" ]; then
  # Take action if $BuildDir doesn‘t exists. 
  echo "create ${BuildDir}..."
  mkdir -p ${BuildDir}
fi
cd ${BuildDir}

echo "building mobilenet quant"
export LD_LIBRARY_PATH=/opt/fsl-imx-xwayland/4.14-sumo/sysroots/aarch64-poky-linux/usr/lib
source /opt/fsl-imx-xwayland/4.14-sumo/environment-setup-aarch64-poky-linux

cmake -DCMAKE_TOOLCHAIN_FILE=../toolchain.cmake -DBUILD_SHARED_LIBS=ON –build –config Release ..

make -j$(nproc)

CMakeLists.txt

cmake_minimum_required(VERSION 2.8.4)  #版本最小为2.8.4
 
PROJECT(mobilenet_quant)   #设置工程名

set(ARMNN_BUILD "/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/armnn-install/lib")
set(ARMNN_ROOT "/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-branches-armnn_20_05")
set(PROTOBUF "/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/google/arm64_pb_install/lib")
set(BOOST_ROOT "/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/boost_arm64_install")
set(FLATBUFFERS "/home/robert/DeepLearning/NXP-imx8QX/CompileResult/armnn-devenv/flatbuffer/arm64_fb_install/lib")

SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++14 -O3 -lpthread -DDLIB_JPEG_SUPPORT -DDLIB_PNG_SUPPORT -DARMNN_TF_LITE_PARSER -fPIE -pie")
SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -I${ARMNN_ROOT}/include -I${ARMNN_ROOT}/src/backends -I${ARMNN_ROOT}/src/armnnUtils -I${ARMNN_ROOT}/tests -I${BOOST_ROOT}/include")
 
IF(CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
  SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Weverything")
ELSEIF(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
  SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra")
ENDIF()

FIND_PACKAGE(OpenCV REQUIRED)
INCLUDE_DIRECTORIES(${OpenCV_INCLUDE_DIRS})
message(STATUS "Opencv include dir found at ${OpenCV_INCLUDE_DIRS}")

find_package(Threads)

set(envSTR "arm")
if(envSTR STREQUAL "x64")
    INCLUDE_DIRECTORIES(/home/robert/DeepLearning/NXP-imx8QX/CompileResult/mydlib-19.19.0)
    LINK_DIRECTORIES(/home/robert/DeepLearning/NXP-imx8QX/CompileResult/mydlib-19.19.0/x64_build_second/dlib/)
ELSEIF(envSTR STREQUAL "arm")
    INCLUDE_DIRECTORIES(/home/robert/DeepLearning/NXP-imx8QX/CompileResult/dlib-19.19.0)
    LINK_DIRECTORIES(/home/robert/DeepLearning/NXP-imx8QX/CompileResult/dlib-19.19.0/dlib_build_second_armv8/dlib/)
    # INCLUDE_DIRECTORIES(${ARMNN_ROOT}/include)
    # INCLUDE_DIRECTORIES(${ARMNN_ROOT}/src/backends)
    # INCLUDE_DIRECTORIES(${ARMNN_ROOT}/src/armnnUtils)
    # INCLUDE_DIRECTORIES(${ARMNN_ROOT}/tests)    
    # INCLUDE_DIRECTORIES(${BOOST_ROOT}/include)
    LINK_DIRECTORIES(${ARMNN_BUILD})
    LINK_DIRECTORIES(${BOOST_ROOT}/lib)
    LINK_DIRECTORIES(${PROTOBUF})
    LINK_DIRECTORIES(${FLATBUFFERS})
ENDIF()

message(${ARMNN_ROOT}/include)

FILE(GLOB SRC_FILES "./*.cpp")
FILE(GLOB HEAD_FILES "./*.h")

ADD_EXECUTABLE(${PROJECT_NAME}  ${SRC_FILES} ${HEAD_FILES})

TARGET_LINK_LIBRARIES(${PROJECT_NAME} dlib armnn armnnTfLiteParser boost_system boost_filesystem boost_program_options protobuf pthread flatbuffers ${OpenCV_LIBS} ${CMAKE_THREAD_LIBS_INIT})

toolchain.cmake

同上

板上测试，针对CPU

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:. ./mobilenet_quant -m ./models/mobilenet_v1_1.0_224_quant.tflite -d ./data/Dog.jpg -p ./models/labels.txt -c CpuAcc

mobilenet_v1_1.0_224_quant.tgz下载地址：https://www.tensorflow.org/lite/guide/hosted_models

labels.txt在源码的/armnn-branches-armnn_20_05/tests/TfLiteMobilenetQuantized-Armnn

Dog.jpg为kaggle猫狗二分类竞赛图片