1 依赖包版本说明
表格中列举了我试验成功的所有依赖包及其对应版本号，部分依赖包可能和官方文档中所推荐的不同。经本人验证，caffe编译过程中所需的依赖包版本相互之间有严重的依赖关系，当然其他版本可能也能正常运行，但如果想节省时间的话，不妨试试以下这些吧~
| Package |
Version |
Description |
| Protobuf |
3.6.0 |
Protocol Buffers - Google's data interchange format. |
| Leveldb |
1.20 |
LevelDB is a fast key-value storage library written at Google that provides an ordered mapping from string keys to string values. |
| Snappy |
1.1.6 |
A fast compressor/decompressor. |
| Opencv |
3.3.1 |
Open Source Computer Vision Library. |
| Boost |
1.56.0 |
Boost provides free peer-reviewed portable C++ source libraries. |
| Atlas |
3.8.4 |
线性代数库 |
| Gflags |
2.1.2 |
The gflags package contains a C++ library that implements commandline flags processing. |
| Glog |
0.3.3 |
Google logging module |
| Lmdb |
0.9.18 |
轻量级内存映射数据库 |
2 依赖包源码编译安装
2.1 Protobuf
./autogen.sh
./configure --prefix=/path/to/install
make
make check
make install
protoc --version #检查是否安装成功-
Python binding安装（这样可以确保caffe和pycaffe使用的是同一版protobuf）
cd $root/python
python setup.py build
python setup.py install
2.2 leveldb
make -j8
cp -r include/* path_to_install/include/
cp out-shared/* path_to_install/lib/2.3 gflags
mkdir build
cd build
export CXXFLAGS="-fPIC”
cmake .. -DBUILD_SHARED_LIBS=ON -DCMAKE_INSTALL_PREFIX=/path/to/install
make VERBOSE=1
make -j8
make install2.4 glog
./configure --prefix=/path/to/install
make -j8
make install2.5 boost
./bootstrap.sh --prefix=path/to/install
./b2 install2.6 opencv
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/path/to/install
make -j8
make install2.7 snappy
./configure --prefix=/path/to/install
make
make install2.8 atlas
- 查阅了多方资料，大家都不太建议自行编译，建议使用apt-get等工具安装
- sudo apt-get install libatlas-base-dev
2.9 lmdb
cd lmdb-xxx/libraries/liblmdb
# 修改29行：prefix=/path/to/install
vim Makefile
make
make install
3 caffe 编译
-
编译—caffe：
# Step1:备份makefile.config --------------
cp Makefile.config.example Makefile.config
# ----------------------------------------
# Step2:根据自己的路径 对makefile.config进行相应修改 ----------------------------------------------
# 因为我们使用的是OpenCV3，所以需要取消注释
# Uncomment if you're using OpenCV 3
OPENCV_VERSION := 3
# 设置CUDA路径，若编译CPU版，则需要打开CPU_ONLY选项
# CUDA directory contains bin/ and lib/ directories that we need.
CUDA_DIR := /usr/local/cuda
# 设置Python头文件路径，主要是Python.h和numpy头文件
# We need to be able to find Python.h and numpy/arrayobject.h.
PYTHON_INCLUDE := /home/conda/include/python2.7/ \
/home/conda/lib/python2.7/site-packages/numpy/core/include
# 设置Python库目录
# We need to be able to find libpythonX.X.so or .dylib.
PYTHON_LIB := /home/conda/lib
# 设置其他依赖包的头文件路径和库目录
# Whatever else you find you need goes here.
INCLUDE_DIRS := $(PYTHON_INCLUDE) /home/local/include
LIBRARY_DIRS := $(PYTHON_LIB) /home/local/lib
#------------------------------------------------------------------------------------------------
# Step3: 编译 --------------
make all -j8
make test
make runtest
#----------------------------
编译—pycaffe
make pycaffe
export PYTHONPATH=$caffe_root/python:$PYTHONPATH4 问题汇总
Q1: 在python中import caffe时，提示"No module named google.protobuf.internal"
A: 卸载掉现在的protobuf，参考2.1重新安装。
Q2: xxx/libstdc++.so.6: version `CXXABI_1.3.9' not found
A: 参考https://blog.csdn.net/zx714311728/article/details/69628836
Q3: ImportError: numpy.core.multiarray failed to import
A: 卸载numpy，重新安装一个版本号>1.14的numpy
Q4: import caffe 时出现Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so
A: 参考https://blog.csdn.net/u010335339/article/details/51501246
Q5: #error This file requires compiler and library support for the ISO C++ 2011 standard. This support is currently experimental, and must be enabled with the -std=c++11 or -std=gnu++11 compiler options.
A: 说明所使用的编译器默认不是C++11标准，需要修改caffe/Makefile相应位置，添上-std=c++11选项。具体有以下几处：
# <caffe/Makefile>
414 CXXFLAGS += -pthread -fPIC -std=c++11 $(COMMON_FLAGS) $(WARNINGS)
415 NVCCFLAGS += -ccbin=$(CXX) -Xcompiler -fPIC -std=c++11 $(COMMON_FLAGS)
507 @ echo CXX/LD -o $@ $<
508 $(Q)$(CXX) -shared -o $@ $(PY$(PROJECT)_SRC) \
509 -o $@ $(LINKFLAGS) -l$(LIBRARY_NAME) $(PYTHON_LDFLAGS) \
510 -Wl,-rpath,$(ORIGIN)/../../build/lib -std=c++11
572 @ echo LD -o $@
573 $(Q)$(CXX) -shared -o $@ $(OBJS) $(VERSIONFLAGS) $(LINKFLAGS) $(LDFLAGS) -std=c++11
603 @ echo CXX/LD -o $@ $<
604 $(Q)$(CXX) $(TEST_MAIN_SRC) $(TEST_OBJS) $(GTEST_OBJ) \
605 -o $@ $(LINKFLAGS) $(LDFLAGS) -l$(LIBRARY_NAME) -Wl,-rpath,$(ORIGIN)/../lib -std=c++11
609 @ echo LD $<
610 $(Q)$(CXX) $(TEST_MAIN_SRC) $< $(GTEST_OBJ) \
611 -o $@ $(LINKFLAGS) $(LDFLAGS) -l$(LIBRARY_NAME) -Wl,-rpath,$(ORIGIN)/../lib -std=c++11
615 @ echo LD $<
616 $(Q)$(CXX) $(TEST_MAIN_SRC) $< $(GTEST_OBJ) \
617 -o $@ $(LINKFLAGS) $(LDFLAGS) -l$(LIBRARY_NAME) -Wl,-rpath,$(ORIGIN)/../lib -std=c++11
625 @ echo CXX/LD -o $@
626 $(Q)$(CXX) $< -o $@ $(LINKFLAGS) -l$(LIBRARY_NAME) $(LDFLAGS) \
627 -Wl,-rpath,$(ORIGIN)/../lib -std=c++11
630 @ echo CXX/LD -o $@
631 $(Q)$(CXX) $< -o $@ $(LINKFLAGS) -l$(LIBRARY_NAME) $(LDFLAGS) \
632 -Wl,-rpath,$(ORIGIN)/../../lib -std=c++11Q6: 在编译MobileNet-Yolo时遇到这样的问题：opencv-3.3.1/lib/libopencv_videoio.so.3.3: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
make: *** [.build_release/examples/ssd/ssd_detect.bin] Error 1
A: 因为Makefile里有个小bug，在使用OpenCV3时少引了一个库，在Makefile的第198行后面添上opencv_videoio即可。
Q7: 编译caffe-ssd时遇到如下问题：/lib/libcaffe.so: undefined reference to `boost::re_detail::cpp_regex_traits_implementation<char>::transform(char constmake: *** [.build_release/tools/upgrade_net_proto_binary.bin] Error 1
*, char const*) const'
A: 在Makefile第181行后面添加boost_regex