不纠结——caffe安装之旅(ubuntu16.10+gcc4.9+python2.7+cuda8.0+opencv3)

本篇仅用来记录我的安装步骤，不一定有普适性。
前提：纯净的ubuntu16.10系统

1.安装显卡驱动

安装什么驱动要看你的显卡是什么，我的是GTX1080Ti，选择了nvidia-384

sudo apt-get remove --purge nvidia*
sudo add-apt-repository ppa:xorg-edgers/ppa
sudo apt-get update
sudo apt-get install nvidia-384
nvidia-smi

2.gcc降级

降级的原因在于系统自带的gcc是6.2版本，cuda不支持

sudo apt-get install gcc-4.9 g++-4.9
cd /usr/bin
sudo rm gcc g++
sudo ln -s gcc-4.9 gcc
sudo ln -s g++-4.9 g++

3.安装cuda V8.0.61

安装时注意不要安装显卡驱动，因为上一步已经安装了！
官网下载cuda_8.0.61_375.26_linux.run和cuda_8.0.61.2_linux.run

sudo sh cuda_8.0.61_375.26_linux.run
sudo sh cuda_8.0.61.2_linux.run
sudo vim ~/.bashrc
# added by cuda-8.0
export PATH=/usr/local/`这里写代码片`cuda-8.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
export CUDA_HOME=/usr/local/cuda
source ~/.bashrc

4.安装cudnn V5.1

官网下载cudnn-8.0-linux-x64-v5.1.tar

tar xvf cudnn-8.0-linux-x64-v5.1.tar
cd cuda
sudo cp include/cudnn.h /usr/local/cuda/include/
sudo cp lib64/lib* /usr/local/cuda/lib64/
cd /usr/local/cuda/lib64/sudo rm -rf libcudnn.so libcudnn.so.5
sudo ln -s libcudnn.so.5.1.10 libcudnn.so.5
sudo ln -s libcudnn.so.5 libcudnn.so

5.安装opencv3

查看当前是否安装了opencv:

pkg-config --modversion opencv 

安装依赖项：

sudo apt-get install git
sudo apt-get install python-pip
pip install cmake
sudo apt-get install libgtk2.0-dev pkg-config python-dev python-numpy
sudo apt-get install libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev 
sudo apt-get install libjasper-dev libdc1394-22-dev

下载opencv3源码：

wget https://github.com/opencv/opencv/archive/3.3.1.zip
unzip 3.3.1.zip
cd opencv-3.3.1/
mkdir build &&cd build
cmake -D CMAKE_BUILD_TYPE=Release -D CMAKE_INSTALL_PREFIX=/usr/local ..
make -j7
sudo make install -j7

6.源码编译protobuf

如果直接apt-get安装很有可能导致错误

sudo apt-get autoremove libprotobuf-dev protobuf-compiler
git clone https://github.com/google/protobuf.git
cd protobuf
./autogen.sh
./configure
makemake check
sudo make install
ldconfig

7.源码编译leveldb

如果直接apt-get安装很有可能导致错误

sudo apt-get autoremove libleveldb-dev
wget https://codeload.github.com/google/leveldb/zip/master
unzip master
cd leveldb-master && make all
sudo cp out-shared/libleveldb.so* /usr/local/lib
sudo cp -R include/* /usr/local/include

8.源码编译glog

如果直接apt-get安装很有可能导致错误

sudo apt-get autoremove libgoogle-glog-dev
wget https://github.com/google/glog/archive/v0.3.3.tar.gz
tar zxvf v0.3.3.tar.gz
cd glog-0.3.3
./configure
make -j7
sudo make install

9.源码编译gflags

如果直接apt-get安装很有可能导致错误

sudo apt-get autoremovelibgflags-dev
wget https://github.com/schuhschuh/gflags/archive/master.zip
unzip master.zipcd gflags-master
mkdir build && cd buildexport CXXFLAGS="-fPIC" && cmake .. && make VERBOSE=1
make -j7
sudo make install

10.源码编译boost

如果直接apt-get安装很有可能导致错误

sudo apt-get install libbz2-dev
sudo apt-get autoremovelibboost-all-dev
wget https://sourceforge.net/projects/boost/files/boost/1.61.0/boost_1_61_0.tar.gz
tar zxvf boost_1_61_0.tar.gz
cd boost_1_61_0
./bootstrap.sh
./b2
sudo ./b2 install

11.安装caffe & pycaffe

安装基本依赖项：

sudo apt-get install libsnappy-dev libopencv-dev libhdf5-serial-dev 
sudo apt-get install libopenblas-dev liblapack-dev libatlas-base-dev
sudo apt-get install liblmdb-dev

下载caffe

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

# 修改如下几处地方,尤其是要把hdf5的路径加进去，否则会找不到
USE_CUDNN := 1
OPENCV_VERSION := 3
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial  
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial

安装pycaffe依赖项：

cd caffe/python
for req in $(cat requirements.txt); do sudo pip install $req; done

编译caffe & pycaffe

make clean
make pycaffe
make all
make test
make runtest

添加环境变量

sudo vim ~/.bashrc
# added by pycaffe
export PYTHONPATH=/home/zhoujie/caffe/python:$PYTHONPATH
source ~/.bashrc

测试：

python
import caffe
# 没有报错信息即安装好了