本文参考caffe官网教程以及网上的两篇教程：Ubuntu14.04+CPU+Python的Caffe安装教程、Caffe学习系列(13)：数据可视化环境（python接口)配置编写而成，因为过程比较波折，记录下来以备日后查用
安装编译caffe的各种依赖
安装基础依赖
1sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler 2sudo apt-get install --no-install-recommends libboost-all-dev
安装BLAS
sudo apt-get install libatlas-base-dev
其他的一些依赖
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev
Python Caffe
安装Python
sudo apt-get install python-dev
###安装anaconda 从清华的镜像下载anaconda，根据anaconda官网提供的版本号，从清华镜像的目录中查找到对应的版本为Anaconda2-5.0.1-Linux-x86_64.sh（python2.7版本）
wget https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda2-5.0.1-Linux-x86_64.sh
下载完成后，运行这个sh。在安装的过程中，会问你安装路径，直接回车默认就可以了。有个地方问你是否将anaconda安装路径加入到环境变量（.bashrc)中，这个一定要输入yes。
安装成功后，会有当前用户根目录下生成一个anaconda2的文件夹，里面就是安装好的内容。
下载Caffe
git clone https://github.com/BVLC/caffe.git
配置python
将caffe根目录下的python文件夹加入到环境变量，等编译好了以后就可以import caffe来使用了
打开配置文件bashrc
sudo vi ~/.bashrc
在最后面加入
export PYTHONPATH=$(caffe_path)/python:$PYTHONPATH
注意 $(caffe_path) 需要根据自己caffe安装路径的实际情况配置
保存退出，更新配置文件
source ~/.bashrc
修改Makefile.config
首先，在caffe的根目录下复制Makefile.config.example 成 Makefile.config
cp Makefile.config.example Makefile.config
然后修改Makefile.config，我的配置如下所示
1## Refer to http://caffe.berkeleyvision.org/installation.html 2# Contributions simplifying and improving our build system are welcome! 3 4# cuDNN acceleration switch (uncomment to build with cuDNN). 5# USE_CUDNN := 1 6 7# CPU-only switch (uncomment to build without GPU support). 8#因为使用的是cpu模式，所以要放开 9 CPU_ONLY := 1 10 11# uncomment to disable IO dependencies and corresponding data layers 12# USE_OPENCV := 0 13# USE_LEVELDB := 0 14# USE_LMDB := 0 15 16# uncomment to allow MDB_NOLOCK when reading LMDB files (only if necessary) 17# You should not set this flag if you will be reading LMDBs with any 18# possibility of simultaneous read and write 19# ALLOW_LMDB_NOLOCK := 1 20 21# Uncomment if you're using OpenCV 3 22# OPENCV_VERSION := 3 23 24# To customize your choice of compiler, uncomment and set the following. 25# N.B. the default for Linux is g++ and the default for OSX is clang++ 26# CUSTOM_CXX := g++ 27 28# CUDA directory contains bin/ and lib/ directories that we need. 29CUDA_DIR := /usr/local/cuda 30# On Ubuntu 14.04, if cuda tools are installed via 31# "sudo apt-get install nvidia-cuda-toolkit" then use this instead: 32# CUDA_DIR := /usr 33 34# CUDA architecture setting: going with all of them. 35# For CUDA < 6.0, comment the *_50 through *_61 lines for compatibility. 36# For CUDA < 8.0, comment the *_60 and *_61 lines for compatibility. 37# For CUDA >= 9.0, comment the *_20 and *_21 lines for compatibility. 38CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \ 39 -gencode arch=compute_20,code=sm_21 \ 40 -gencode arch=compute_30,code=sm_30 \ 41 -gencode arch=compute_35,code=sm_35 \ 42 -gencode arch=compute_50,code=sm_50 \ 43 -gencode arch=compute_52,code=sm_52 \ 44 -gencode arch=compute_60,code=sm_60 \ 45 -gencode arch=compute_61,code=sm_61 \ 46 -gencode arch=compute_61,code=compute_61 47 48# BLAS choice: 49# atlas for ATLAS (default) 50# mkl for MKL 51# open for OpenBlas 52BLAS := atlas 53# Custom (MKL/ATLAS/OpenBLAS) include and lib directories. 54# Leave commented to accept the defaults for your choice of BLAS 55# (which should work)! 56# BLAS_INCLUDE := /path/to/your/blas 57# BLAS_LIB := /path/to/your/blas 58 59# Homebrew puts openblas in a directory that is not on the standard search path 60# BLAS_INCLUDE := $(shell brew --prefix openblas)/include 61# BLAS_LIB := $(shell brew --prefix openblas)/lib 62 63# This is required only if you will compile the matlab interface. 64# MATLAB directory should contain the mex binary in /bin. 65# MATLAB_DIR := /usr/local 66# MATLAB_DIR := /Applications/MATLAB_R2012b.app 67 68# NOTE: this is required only if you will compile the python interface. 69# We need to be able to find Python.h and numpy/arrayobject.h. 70#PYTHON_INCLUDE := /usr/include/python2.7 \ 71 /usr/lib/python2.7/dist-packages/numpy/core/include 72# Anaconda Python distribution is quite popular. Include path: 73# Verify anaconda location, sometimes it's in root. 74#因为使用了anaconda所以要配置anaconda的地址，如果不配置则会出现找不到*.h的情况 75 ANACONDA_HOME := $(HOME)/anaconda2 76#使用anaconda的头文件 77 PYTHON_INCLUDE := $(ANACONDA_HOME)/include \ 78 $(ANACONDA_HOME)/include/python2.7 \ 79 $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include 80 81# Uncomment to use Python 3 (default is Python 2) 82# PYTHON_LIBRARIES := boost_python3 python3.5m 83# PYTHON_INCLUDE := /usr/include/python3.5m \ 84# /usr/lib/python3.5/dist-packages/numpy/core/include 85 86# We need to be able to find libpythonX.X.so or .dylib. 87# PYTHON_LIB := /usr/lib 88#使用anaconda的库 89PYTHON_LIB := $(ANACONDA_HOME)/lib 90 91# Homebrew installs numpy in a non standard path (keg only) 92# PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; print(numpy.core.__file__)'))/include 93# PYTHON_LIB += $(shell brew --prefix numpy)/lib 94 95# Uncomment to support layers written in Python (will link against Python libs) 96# WITH_PYTHON_LAYER := 1 97 98# Whatever else you find you need goes here. 99INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include 100LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib 101 102# If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies 103# INCLUDE_DIRS += $(shell brew --prefix)/include 104# LIBRARY_DIRS += $(shell brew --prefix)/lib 105 106# NCCL acceleration switch (uncomment to build with NCCL) 107# https://github.com/NVIDIA/nccl (last tested version: v1.2.3-1+cuda8.0) 108# USE_NCCL := 1 109 110# Uncomment to use `pkg-config` to specify OpenCV library paths. 111# (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.) 112# USE_PKG_CONFIG := 1 113 114# N.B. both build and distribute dirs are cleared on `make clean` 115BUILD_DIR := build 116DISTRIBUTE_DIR := distribute 117 118# Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171 119# DEBUG := 1 120 121# The ID of the GPU that 'make runtest' will use to run unit tests. 122TEST_GPUID := 0 123 124# enable pretty build (comment to see full commands) 125Q ?= @
编译caffe
1make pycaffe -j8 2make all -j8 3make test -j8 4make runtest -j8
运行python
进入caffe/python ，运行python
1python 2Python 2.7.14 |Anaconda custom (64-bit)| (default, Oct 16 2017, 17:29:19) 3[GCC 7.2.0] on linux2 4Type "help", "copyright", "credits" or "license" for more information. 5>>> import caffe 6>>>
输入import caffe没有反应则说明成功，但如果出现问题error ：No module named google.protobuf.internal，则运行以下代码：
conda install protobuf
接下来，运行caffe自带的例子，可参考运行caffe自带的两个简单例子