ARM鲲鹏920-oe2309-caffe

参考链接:Caffe | Installation

安装依赖包

dnf install

dnf update
dnf install leveldb-devel snappy-devel opencv.aarch64 boost-devel hdf5-devel gflags-devel glog-devel lmdb-devel openblas.aarch64
dnf install git wget tar gcc-g++ unzip automake libtool autoconf

从源码编译安装protobuf

问题：

编译caffe需要用到protobuf，通过dnf install protobuf-devel可以获得3.19版本，但在后续编译caffe时会出现下图报错。报错原因：https://github.com/onnx/onnx/issues/2678，故而需要安装3.6~3.10版本的protobuf。

安装：

git clone https://github.com/protocolbuffers/protobuf.git
cd protobuf
git checkout 3.9.x
./autogen.sh
./configure --prefix=/home/caffe/file/protobuf-install
make -j16
make install -j16
export PROTOBUF="/home/caffe/file/protobuf-install"
export PATH="$PROTOBUF/bin:$PATH"
export LD_LIBRARY_PATH="$PROTOBUF/lib:$LD_LIBRARY_PATH"
protoc --version

如果出现了版本号，则证明安装成功

安装caffe

git clone https://github.com/BVLC/caffe.git
cd caffe
cp Makefile.config.example Makefile.config
vi Makefile.config

1. 选择cpu-only模式：将第8行CPU_ONLY := 1取消注释
2. 选择BLAS：将第53行BLAS := atlas改为BLAS := open
3. 修改INCLUDE路径：将第97行INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include改为INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/opencv4 /usr/include/openblas /home/caffe/file/protobuf-install/include
4. 修改LIBRARY路径：将第98行LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib改为LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /home/caffe/file/protobuf-install/lib
5. 添加LIBRARIES：在第98行后面添加一行LIBRARIES += opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs opencv_videoio

在caffe路径下运行下面命令：

sed -i 's/CV_LOAD_IMAGE_COLOR/cv::IMREAD_COLOR/g' src/caffe/layers/window_data_layer.cpp
sed -i 's/CV_LOAD_IMAGE_COLOR/cv::IMREAD_COLOR/g' src/caffe/util/io.cpp
sed -i 's/CV_LOAD_IMAGE_GRAYSCALE/cv::ImreadModes::IMREAD_GRAYSCALE/g' src/caffe/util/io.cpp
sed -i 's/CV_LOAD_IMAGE_COLOR/cv::IMREAD_COLOR/g' src/caffe/test/test_io.cpp
sed -i 's/CV_LOAD_IMAGE_GRAYSCALE/cv::ImreadModes::IMREAD_GRAYSCALE/g' src/caffe/test/test_io.cpp
make all -j16
make test -j16
make runtest -j16
export CAFFE="/home/caffe/file/caffe"
export PATH="$CAFFE/build/tools:$PATH"
caffe --version

如果出现了版本号，则证明安装成功