Tengine Android版本编译

环境

System: Ubuntu18.04
CMAKE: 3.15.3

下载Android NDK

wget ftp://ftp.openailab.net.cn/Tengine_android_build/android-ndk-r16b-linux-x86_64.zip
unzip android-ndk-r16b-linux-x86_64.zip

克隆Tengine源码

git clone https://github.com/OAID/Tengine.git

编译

设置ANDROID_NDK环境变量

ANDROID_NDK为刚才解压的zip文件

export ANDROID_NDK=/home/oal/ssd_data/workspace/tmp/android-ndk-r16b

编译ARM32

cd Tengine
mkdir build_android_32
cd build_android_32
cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_PLATFORM=android-22 -DANDROID_STL=c++_shared -DANDROID_ARM_NEON=ON -DCONFIG_ARCH_ARM32=ON -DANDROID_ALLOW_UNDEFINED_SYMBOLS=TRUE ..

如果输出下面内容，说明没有问题

-- Configuring done
-- Generating done

编译

make -j4 && make install

编译ARM64

cd Tengine
mkdir build_android_64
cd build_android_64
cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-22 -DANDROID_STL=c++_shared -DANDROID_ARM_NEON=ON -DCONFIG_ARCH_ARM64=ON -DANDROID_ALLOW_UNDEFINED_SYMBOLS=TRUE ..

如果输出下面内容，说明没有问题

-- Configuring done
-- Generating done

编译

make -j4 && make install

测试

绝大多数手机都支持ARM32，我们用编译好的ARM32版本测试编译出来的so有没有问题。

把 https://github.com/jiangzhong... 中的cat.jpg，squeezenet_caffe.tmfile放到Tengine/build_android_32/install/bin目录下

然后push到手机的/data/local/tmp/目录下，这个tmp目录是Android系统的特殊目录，放在里面的文件可以被赋予可执行权限。

cd Tengine/build_android_32/install
adb push ./lib/libtengine-lite.so /data/local/tmp/
adb push ./bin/tm_classification /data/local/tmp/
adb push ./bin/cat.jpg /data/local/tmp/
adb push ./bin/squeezenet_caffe.tmfile /data/local/tmp/

有些手机可能无法直接push到/data/local/tmp文件夹，这种情况下可以先push到/sdcard/文件夹，然后在mv到/data/local/tmp/

用下面命令进入手机终端并进行测试

adb shell
cd /data/local/tmp/
export LD_LIBRARY_PATH=.
chmod +x tm_classification
./tm_classification -m squeezenet_caffe.tmfile -i cat.jpg

输出

Image height not specified, use default 227
Image width not specified, use default  227
Scale value not specified, use default  1.0, 1.0, 1.0
Mean value not specified, use default   104.0, 116.7, 122.7
tengine-lite library version: 0.2-dev

model file : squeezenet_caffe.tmfile
image file : ./cat.jpg
img_h, img_w, scale[3], mean[3] : 227 227 , 1.000 1.000 1.000, 104.0 116.7 122.7
Repeat 1 times, thread 1, avg time 33.31 ms, max_time 33.31 ms, min_time 33.31 ms
--------------------------------------
0.273201, 281
0.267551, 282
0.181003, 278
0.081799, 285
0.072407, 151
--------------------------------------

自此Tengine Android版本编译成功。

自动化编译脚本

wget ftp://ftp.openailab.net.cn/Tengine_android_build/android-ndk-r16b-linux-x86_64.zip
unzip android-ndk-r16b-linux-x86_64.zip
export ANDROID_NDK=`pwd`/android-ndk-r16b


git clone https://github.com/OAID/Tengine.git
git clone https://github.com/jiangzhongbo/Tengine_Tutorial
cd Tengine

mkdir build_android_32
cd build_android_32
cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI="armeabi-v7a" -DANDROID_PLATFORM=android-22 -DANDROID_STL=c++_shared -DANDROID_ARM_NEON=ON -DCONFIG_ARCH_ARM32=ON -DANDROID_ALLOW_UNDEFINED_SYMBOLS=TRUE ..
make -j4 && make install

cd ./install/bin
cp ../../../../Tengine_Tutorial/0_Compile/cat.jpg ./
cp ../../../../Tengine_Tutorial/0_Compile/squeezenet_caffe.tmfile ./

cd ../../..

mkdir build_android_64
cd build_android_64
cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI="arm64-v8a" -DANDROID_PLATFORM=android-22 -DANDROID_STL=c++_shared -DANDROID_ARM_NEON=ON -DCONFIG_ARCH_ARM64=ON -DANDROID_ALLOW_UNDEFINED_SYMBOLS=TRUE ..
make -j4 && make install

cd ./install/bin
cp ../../../../Tengine_Tutorial/0_Compile/cat.jpg ./
cp ../../../../Tengine_Tutorial/0_Compile/squeezenet_caffe.tmfile ./

为了方便，你可以直接

wget https://raw.githubusercontent.com/jiangzhongbo/Tengine_Tutorial/master/0_Compile/compile_tengine_android.sh
sh compile_tengine_android.sh

参考

https://github.com/OAID/Tengine/wiki/Tengine源码编译

源码

https://github.com/jiangzhongbo/Tengine_Tutorial

知乎

https://zhuanlan.zhihu.com/p/182743221

更多Tengine相关内容请关注Tengine-边缘AI推理框架专栏。