[关闭]
@Wayne-Z 2018-01-17T19:16:43.000000Z 字数 4268 阅读 2698

Caffe Ubuntu16.04环境配置

Caffe



环境

官网最新版Ubuntu16.04.3

CPU-Only

依照教程,执行如下过程

  1. sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler
  2. sudo apt-get install --no-install-recommends libboost-all-dev

安装Anaconda的Python3.6 version
强烈推荐vscode进行文本编辑。

补充依赖包

网盘中有glog包,下载后

  1. tar zxvf glog-0.3.3.tar.gz
  2. cd glog-0.3.3
  3. ./configure
  4. sudo make
  5. sudo make install

同时,依据Ubuntu16.04安装caffe中遇到的问题总结,命令行输入如下命令继续安装相关依赖包

  1. #1. ./include/caffe/common.hpp:5:27: fatal error: gflags/gflags.h: No such file or directory
  2. 解决办法:sudo apt-get install libgflags-dev
  3. #2. ./include/caffe/util/mkl_alternate.hpp:14:19: fatal error: cblas.h: No such file or directory
  4. 解决办法:sudo apt-get install libblas-dev
  5. #3. ./include/caffe/util/db_lmdb.hpp:8:18: fatal error: lmdb.h: No such file or directory
  6. 解决办法:sudo apt install liblmdb-dev
  7. #4. /usr/bin/ld: cannot find -lcblas
  8. /usr/bin/ld: cannot find -latlas
  9. 解决办法:sudo apt install libatlas-base-dev

其中关于HDF5的问题,不是按照博客原来的方式解决的,博客原来采用这种方式

  1. #3. ./include/caffe/util/hdf5.hpp:6:18: fatal error: hdf5.h: No such file or directory
  2. 解决办法:在Makefile.config找到以下行并添加蓝色部分
  3. INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial
  4. LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial

但是这个在进入后面make runtest的时候会出现HDF5的相关问题,无法通过增加软链接来解决。这里参照了另一个办法并进行了细微的修改,去官网下载
下载解压后,输入命令

  1. ./configure --prefix=/usr/local/lib/hdf5-1.8.3/
  2. make ##4分钟左右
  3. sudo make install

这样就直接把hdf5放入了lib的常用位置。但是这个还是跑不通测试,最终在这里找到了办法,anaconda3里面自带了这个库,所以只用添加进去就好。

编译

进入caffe文件夹,输入

  1. cp Makefile.config.example Makefile.config
  2. code Makefile.config

对文件进行了一些修改,针对CPU模式,主要是Anaconda Python的一些参数调整

  1. # CPU-only switch (uncomment to build without GPU support).
  2. CPU_ONLY := 1
  3. # Uncomment if you're using OpenCV 3
  4. # OPENCV_VERSION := 3
  5. # To customize your choice of compiler, uncomment and set the following.
  6. # N.B. the default for Linux is g++ and the default for OSX is clang++
  7. CUSTOM_CXX := g++
  8. # NOTE: this is required only if you will compile the python interface.
  9. # We need to be able to find Python.h and numpy/arrayobject.h.
  10. # PYTHON_INCLUDE := /usr/include/python2.7 \
  11. #/usr/lib/python2.7/dist-packages/numpy/core/include
  12. # Anaconda Python distribution is quite popular. Include path:
  13. # Verify anaconda location, sometimes it's in root.
  14. ANACONDA_HOME := $(HOME)/anaconda3
  15. PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
  16. $(ANACONDA_HOME)/include/python3.6m \
  17. $(ANACONDA_HOME)/lib/python3.6/site-packages/numpy/core/include
  18. # We need to be able to find libpythonX.X.so or .dylib.
  19. # PYTHON_LIB := /usr/lib
  20. PYTHON_LIB := $(ANACONDA_HOME)/lib
  21. # Uncomment to support layers written in Python (will link against Python libs)
  22. WITH_PYTHON_LAYER := 1
  23. # Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171
  24. DEBUG := 1

之后就是

  1. make all
  2. make test
  3. make runtest

可以全部通过,没有问题。

安装Opencv

最核心的部分安装成功后,安装opencv支持。

  1. sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev

百度网盘下载准备好opencv和opencv_contrib两个文件夹,而后

  1. cd opencv
  2. mkdir build
  3. cd build
  4. cmake D CMAKE_BUILD_TYPE=Release D OPENCV_EXTRA_MODULES_PATH=/home/banshee/opencv_contrib/modules/ -D CMAKE_INSTALL_PREFIX=/usr/local D WITH_IPP=OFF ..

在这个过程中,会遇到ippicv_linux_20151201.tgz国内被墙了,有时候下载不成功或者提出hash值不对
image_1c1mptaiaglm16on1s1b3i12rq9.png-755.3kB
此时果断中止,然后下载对应文件到对应目录即可(/3rdparty/ippicv/downloads/linux-808b791a6eac9ed78d32a7666804320e/ippicv_linux_20151201.tgz)
而后

  1. (<opencv目录>/3rdparty/ippicv/downloads/linux-808b791a6eac9ed78d32a7666804320e/ippicv_linux_20151201.tgz

而后

  1. sudo make -j4
  2. sudo make install

然后运行测试代码

  1. #include <iostream>
  2. #include <opencv2/highgui/highgui.hpp>
  3. int main(int argc,char* argv[]){
  4. const std::string window_name = "lena";
  5. const std::string input_pic = "lena.jpg";
  6. cv::Mat test_pic = cv::imread(input_pic);
  7. if(test_pic.empty()){
  8. std::cout<<"no input image"<<std::endl;
  9. return 1;
  10. }
  11. cv::namedWindow(window_name);
  12. cv::imshow(window_name,test_pic);
  13. cv::waitKey(0);
  14. return 0;
  15. }

对应的测试图片lena.jpg在网盘中也有。

  1. g++ -o test_opencv test_opencv.cpp -lopencv_core -lopencv_highgui -lopencv_imgcodecs

完成后,重新编译caffe

  1. make all
  2. make test
  3. make runtest
  4. make pycaffe

Pycaffe

在上文编译好pycaffe后,我们会发现pycaffe并不能使用,会出现如下报错

  1. ImportError: /home/zero/Documents/caffe-master/python/caffe/_caffe.so:undefined symbol:
  2. _ZN5boost6python6detail11init_moduleER11PyModuleDefPFvvE

这个bug是由boost版本不匹配导致的,涉及到Makefile中的这样一行

  1. PYTHON_LIBRARIES ?= boost_python python2.7

为了方便,笔者将它修改成了

  1. PYTHON_LIBRARIES := boost_python3 python3.5m

并且增加了如下操作:

  1. cd /usr/lib/x86_64-linux-gnu/
  2. sudo ln -s libboost_python-py35.so libboost_python3.so
  3. sudo ln -s libpython3.5m.so.1.0 libpython3.5m.so

后面才发现,原来在Makefile.config里面就有一模一样的这句话,可见其实只要取消其中的注释,使用上面三行命令并重新编译即可。

GPU

添加新批注
在作者公开此批注前,只有你和作者可见。
回复批注