Caffe 环境搭建中应注意的问题

和TensorFlow对应的是Theano，Torch； 
Caffe专精于图像处理，Caffe方便，更快入门上手； 
在通用的DL task上，Caffe不如Theano。 
CNN(卷积神经网络)、RNN(循环神经网络)、DNN(深度神经网络)

开发环境搭建：

一、没有GPU

learning@learning-virtual-machine:~$ lspci | grep -i nvidia 
learning@learning-virtual-machine:~$ 
 
 
1
2

二、ubuntu版本

learning@learning-virtual-machine:~$ uname -m && cat /etc/*release
x86_64
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=15.10
DISTRIB_CODENAME=wily
DISTRIB_DESCRIPTION="Ubuntu 15.10"
NAME="Ubuntu"
VERSION="15.10 (Wily Werewolf)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 15.10"
VERSION_ID="15.10"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
learning@learning-virtual-machine:~$ 

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

三、gcc

learning@learning-virtual-machine:~$ gcc --version  
gcc.real (Ubuntu 5.2.1-22ubuntu2) 5.2.1 20151010
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

learning@learning-virtual-machine:~$ 

 
 
1
2
3
4
5
6
7
8

四、安装依赖库

learning@learning-virtual-machine:~$ sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler
[sudo] password for learning: 
Reading package lists... Done
Building dependency tree      
 
 
1
2
3
4

sudo apt-get install --no-install-recommends libboost-all-dev
 
 
1

sudo apt-get install libatlas-base-dev
 
 
1

五、安装python

六、安装Opencv

安装Opencv 
http://blog.csdn.net/forest_world/article/details/51372703Ubuntu

七、安装依赖库 
learning@learning-virtual-machine:~$ sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev 
[sudo] password for learning: 
Reading package lists… Done

八、下载Caffe

learning@learning-virtual-machine:~$ git clone git://github.com/BVLC/caffe.git
Cloning into 'caffe'...
remote: Counting objects: 34637, done.
Receiving objects: 100% (34637/34637), 47.81 MiB | 81.00 KiB/s, done.
remote: Total 34637 (delta 0), reused 0 (delta 0), pack-reused 34636
Resolving deltas: 100% (23287/23287), done.
Checking connectivity... done.

 
 
1
2
3
4
5
6
7
8

九、修改

Makefile 修改： 
LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_serial_hl hdf5_serial opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs

这一块代码不需要修改 

修改处： 
caffe/examples/cpp_classification/classification.cpp文件

十、编译

learning@learning-virtual-machine:~/caffe$ cp Makefile.config.example Makefile.config

learning@learning-virtual-machine:~/caffe$ ls
caffe.cloc       data      INSTALL.md               matlab     src
cmake            docker    LICENSE                  models     tools
CMakeLists.txt   docs      Makefile                 python
CONTRIBUTING.md  examples  Makefile.config          README.md
CONTRIBUTORS.md  include   Makefile.config.example  scripts
learning@learning-virtual-machine:~/caffe$ 
 
 
1
2
3
4
5
6
7
8
9

learning@learning-virtual-machine:~/caffe$ gedit Makefile.config

learning@learning-virtual-machine:~/caffe$ make all 
出现问题：

learning@learning-virtual-machine:~/caffe$ make all
PROTOC src/caffe/proto/caffe.proto
CXX .build_release/src/caffe/proto/caffe.pb.cc
CXX src/caffe/data_transformer.cpp
CXX src/caffe/common.cpp
CXX src/caffe/internal_thread.cpp
CXX src/caffe/blob.cpp
CXX src/caffe/data_reader.cpp
CXX src/caffe/parallel.cpp
CXX src/caffe/util/hdf5.cpp
In file included from src/caffe/util/hdf5.cpp:1:0:
./include/caffe/util/hdf5.hpp:6:18: fatal error: hdf5.h: No such file or directory
compilation terminated.
Makefile:572: recipe for target '.build_release/src/caffe/util/hdf5.o' failed
make: *** [.build_release/src/caffe/util/hdf5.o] Error 1
learning@learning-virtual-machine:~/caffe$ 

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

解决： 
Makefile.config 
INCLUDE_DIRS 
/usr/include/hdf5/serial/ 
Makefile 
LIBRARIES 
hdf5_hl and hdf5 改为 hdf5_serial_hl ，hdf5_serial

出现问题：

LD -o .build_release/lib/libcaffe.so.1.0.0-rc3
CXX tools/finetune_net.cpp
CXX/LD -o .build_release/tools/finetune_net.bin
CXX tools/net_speed_benchmark.cpp
CXX/LD -o .build_release/tools/net_speed_benchmark.bin
CXX tools/compute_image_mean.cpp
CXX/LD -o .build_release/tools/compute_image_mean.bin
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(cv::String const&, int)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(cv::String const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imdecode(cv::_InputArray const&, int)'
collect2: error: ld returned 1 exit status
Makefile:616: recipe for target '.build_release/tools/compute_image_mean.bin' failed
make: *** [.build_release/tools/compute_image_mean.bin] Error 1

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14

解决方法： 
Makefile 
修改： 
LIBRARIES += glog gflags protobuf boost_system boost_filesystem m hdf5_serial_hl hdf5_serial opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs

编译成功：

make test

make runtest

[----------] 2 tests from BatchReindexLayerTest/0, where TypeParam = caffe::CPUDevice<float>
[ RUN      ] BatchReindexLayerTest/0.TestForward
[       OK ] BatchReindexLayerTest/0.TestForward (0 ms)
[ RUN      ] BatchReindexLayerTest/0.TestGradient
[       OK ] BatchReindexLayerTest/0.TestGradient (373 ms)
[----------] 2 tests from BatchReindexLayerTest/0 (374 ms total)

[----------] Global test environment tear-down
[==========] 1058 tests from 146 test cases ran. (134225 ms total)
[  PASSED  ] 1058 tests.
learning@learning-virtual-machine:~/caffe$ 

 
 
1
2
3
4
5
6
7
8
9
10
11
12

十一、配置pycaffe

sudo apt-get install python-numpy python-scipy python-matplotlib python-sklearn python-skimage python-h5py python-protobuf python-leveldb python-networkx python-nose python-pandas python-gflags Cython ipython

sudo apt-get install protobuf-c-compiler protobuf-compiler

learning@learning-virtual-machine:~/caffe$ make pycaffe

learning@learning-virtual-machine:~/caffe$ make pycaffe
CXX/LD -o python/caffe/_caffe.so python/caffe/_caffe.cpp
touch python/caffe/proto/__init__.py
PROTOC (python) src/caffe/proto/caffe.proto
learning@learning-virtual-machine:~/caffe$ 

 
 
1
2
3
4
5
6

sudo gedit /etc/profile 
末尾添加： export PYTHONPATH=/path/to/caffe/python:$PYTHONPATH 
用完整路径 
source /etc/profile

learning@learning-virtual-machine:~/caffe$ python 
Python 2.7.10 (default, Oct 14 2015, 16:09:02) 
[GCC 5.2.1 20151010] on linux2 
Type “help”, “copyright”, “credits” or “license” for more information. 
.>>>

出现问题：

.>>> import caffe
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
ImportError: No module named caffe
 
 
1
2
3
4

解决方法： 
sudo gedit /etc/profile 
export PYTHONPATH=$PYTHONPATH:/home/learning/caffe/python 
source /etc/profile

补充:baidu解释 
Python（英国发音：/ˈpaɪθən/ 美国发音：/ˈpaɪθɑːn/）, 是一种面向对象、解释型计算机程序设计语言，由Guido van Rossum于1989年发明，第一个公开发行版发行于1991年。 
Python是纯粹的自由软件， 源代码和解释器CPython遵循 GPL(GNU General Public License)协议[1] 。 
Python语法简洁清晰，特色之一是强制用空白符(white space)作为语句缩进。 
Python具有丰富和强大的库。它常被昵称为胶水语言，能够把用其他语言制作的各种模块（尤其是C/C++）很轻松地联结在一起。常见的一种应用情形是，使用Python快速生成程序的原型（有时甚至是程序的最终界面），然后对其中[2] 有特别要求的部分，用更合适的语言改写，比如3D游戏中的图形渲染模块，性能要求特别高，就可以用C/C++重写，而后封装为Python可以调用的扩展类库。需要注意的是在您使用扩展类库时可能需要考虑平台问题，某些可能不提供跨平台的实现。