【CPU + GPU版】Ubuntu16.04 + caffe + Anaconda2 + OpenCV3.x（附：问题集锦+测试案例）

【UpdateTime：20180105 - 12：30】



【CPU版】Ubuntu + caffe + Anaconda2 + OpenCV3.x

2017.06.02：整理出第一个版本。

2018.01.05：整理libhdf5_hl.so.100和libopencv_highgui.so等两个问题。

2018.01.05：Ubuntu 16.04中使用Matlab R2016b编译Caffe接口出错的解决方法。



一、本文目的



本文基于上一篇博文《Ubuntu 常用软件安装：Matlab + XX-net + ... + TeamViewer + Tensorflow + Caffe》中的caffe环境的搭建部分进行完善，更详细的进行整理，主要贡献如下：

1、根据本人实践过程进行记录，更加完善记录整个“caffe环境搭建”的过程。

2、归纳总结问题集锦：caffe环境的搭建过程确实很多细节需要处理，本文尽可能收集一些相关的参考资料。

3、本文会尽可能详细的列出每个步骤，图文结合，并附上参考网址。

4、贴出【CPU版本】和【GPU版本】的Makefile.configure配置文件。

5、总结归纳测试案例。

参考网址：



1）caffe的配置过程：http://blog.csdn.net/brightming/article/details/51106629

2）【深度学习】 Ubuntu16.04 caffe Opencv2.4.13 GPU环境配置：http://blog.csdn.net/sinat_17196995/article/details/53466524

3）Ubuntu 常用软件安装：Matlab + XX-net + ... + TeamViewer + Tensorflow + Caffe

4）caffe安装：http://blog.csdn.net/lanxuecc/article/details/51997919

新增参考网址（20180105）

3D - caffe：https://github.com/faustomilletari/3D-Caffe.git

深度学习之常用软件（ubuntu）：http://blog.csdn.net/houchaoqun_xmu/article/details/78869052

发表过CVPR-2017的作者的博客：http://zongweiz.leanote.com/post/Install-caffe-in-Ubuntu

cuda-8.0 + cudnn-5.1 的安装步骤请参考这里：http://blog.csdn.net/houchaoqun_xmu/article/details/78869052

OpenCV-3.3.0（Ubuntu-16.06）请参考这里：http://blog.csdn.net/houchaoqun_xmu/article/details/78565

二、G[b]PU版caffe配置过程[/b]



1、更新安装必要的软件包，去除不必要的包

sudo apt-get update
sudo apt-get upgrade
sudo apt-get install build-essential
sudo apt-get autoremove

2、查看系统GPU信息：lspci | grep -i nvidia

hcq@hcq-To-be-filled-by-O-E-M:~$ lspci | grep -i nvidia
01:00.0 VGA compatible controller: NVIDIA Corporation GM206 [GeForce GTX 960] (rev a1)
01:00.1 Audio device: NVIDIA Corporation Device 0fba (rev a1)

注：如果当前系统尚未安装NVIDIA驱动，需先进行安装，可参考这里。

3、安装并配置CUDA8.0

1）下载对应版本的CUDA安装包（.deb文件）：https://developer.nvidia.com/cuda-downloads

2）cd到文件所在路径，并执行如下命令：

3）sudo dpkg -i cuda-repo-<distro>_<version>_<architecture>.deb 

4）sudo apt-get update 

5）sudo apt-get install cuda

=== 配置环境变量 ===

1）export PATH=“/usr/local/cuda-8.0/bin:$PATH”

2）export LD_LIBRARY_PATH=”/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH“

3）配置环境变量和动态链接库，在命令行输入【sudo gedit /etc/profile】在文件末尾加入【export PATH = “/usr/local/cuda/bin:$PATH”】，保存后关闭文件。

4）保存之后，创建链接文件，输入【sudo gedit /etc/ld.so.conf.d/cuda.conf】在打开的文件中添加如下语句【/usr/local/cuda/lib64】执行【sudo ldconfig】。

5）查看显卡的驱动版本: cat /proc/driver/nvidia/version

6）查看nvcc编译器的版本: nvcc -V

=== 编译CUDA Sample（可选） ===

1）参考网址：http://blog.csdn.net/sinat_17196995/article/details/53466524

2）【cd /usr/local/cuda-8.0/samples/bin/x86_64/linux/release】，然后执行【sudo ./deviceQuery】，出现如下信息，表示成功：

sudo ./deviceQuery
[sudo] hcq 的密码：
./deviceQuery Starting...

CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce GTX 960"
CUDA Driver Version / Runtime Version          8.0 / 8.0
CUDA Capability Major/Minor version number:    5.2
Total amount of global memory:                 1993 MBytes (2090270720 bytes)
( 8) Multiprocessors, (128) CUDA Cores/MP:     1024 CUDA Cores
GPU Max Clock rate:                            1178 MHz (1.18 GHz)
Memory Clock rate:                             3505 Mhz
Memory Bus Width:                              128-bit
L2 Cache Size:                                 1048576 bytes
Maximum Texture Dimension Size (x,y,z)         1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
Maximum Layered 1D Texture Size, (num) layers  1D=(16384), 2048 layers
Maximum Layered 2D Texture Size, (num) layers  2D=(16384, 16384), 2048 layers
Total amount of constant memory:               65536 bytes
Total amount of shared memory per block:       49152 bytes
Total number of registers available per block: 65536
Warp size:                                     32
Maximum number of threads per multiprocessor:  2048
Maximum number of threads per block:           1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch:                          2147483647 bytes
Texture alignment:                             512 bytes
Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
Run time limit on kernels:                     Yes
Integrated GPU sharing Host Memory:            No
Support host page-locked memory mapping:       Yes
Alignment requirement for Surfaces:            Yes
Device has ECC support:                        Disabled
Device supports Unified Addressing (UVA):      Yes
Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = GeForce GTX 960
Result = PASS

4、安装Anaconda2

1）到官网下载安装包（此处选择for Linux）：https://www.continuum.io/downloads/

2）cd到安装包文件所在的目录下并执行：bash Anaconda2-4.3.1-Linux-x86_64.sh

3）配置环境变量：

== sudo gedit ~/.bashrc

== export PATH="/home/hcq/anaconda2/bin:$PATH"
== source ~/.bashrc

5、安装、编译OpenCV3.x

1）sudo git clone https://github.com/jayrambhia/Install-OpenCV

2）在Install-OpenCV/Ubuntu目录下：

== 先执行命令：sudo chmod +x *

== 然后执行sudo ./opencv_latest.sh（此过程需要一段时间，最后出现OpenCV 3.2.0 ready to be used表示成功）

3）cd /home/hcq/Install-OpenCV/Ubuntu/OpenCV/opencv-3.2.0

4）sudo mkdir release

5）cd release

6）sudo cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local ..

7）sudo make（编译过程需要一段时间，要到100%才算是是成功）

8）sudo make install

编译完，环境也自动加上了。

===== 配置文件 =====

1）sudo gedit /etc/ld.so.conf.d/opencv.conf #配置相关文件

2）复制如下代码加入到文件

PKG_CONFIG_PATH=$PKG_CONFIG_PATH:/usr/local/lib/pkgconfig 

export PKG_CONFIG_PATH

3）source /etc/bash.bashrc

6、安装依赖项

1）安装google-glog：参考这里

== 下载地址：https://pan.baidu.com/s/1slpjjeH 
== 将glog‐0.3.3.tar.gz这个文件拷贝到主文件夹下，即/home/yourname 文件夹下，点击鼠标右键，选择提取到此处（也就是解压该文件）解压后在终端输入如下命令：

cd /home/yourname/glog-0.3.3
./configure
make
sudo make install

2）安装其他依赖项：

sudo apt-get install libatlas-base-dev libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev
sudo apt-get install libboost-all-dev libhdf5-serial-dev libgflags-dev libgoogle-glog-dev liblmdb-dev protobuf-compiler

3）安装OpenBLAS

git clone https://github.com/xianyi/OpenBLAS.git cd OpenBLAS
make -j $(($(nproc) + 1))
make install

参考网址：https://github.com/xianyi/OpenBLAS#set-the-number-of-threads-with-environment-variables

注：如果在sudo make runtest 的时候出现错误【libopenblas.so.0: cannot open shared object file: No such file or directory】，第四部分有介绍解决方案。

7、安装其他所需工具（可能导致错误的一些安装，建议读者先跳过本节的安装，遇到问题后再进行安装，印象会比较深刻）

1）conda install libgcc    参考网址：http://blog.csdn.net/u010734277/article/details/70195043

2）sudo apt install ffmpeg     参考网址：http://f.dataguru.cn/thread-738610-1-1.html

8、安装并配置CUDNN（一定要看好对应的版本！！否则后面一些链接库会出错）：https://developer.nvidia.com/rdp/cudnn-download

====================== cudnn-5.1-8.0 ========================

1）切换到安装包所在的目录下：sudo tar xvf cudnn-8.0-linux-x64-v5.1.tgz

2）切换到 cuda/lib64 目录下：sudo cp lib* /usr/local/cuda-8.0/lib64/

3）切换到 cuda/include 目录下：sudo cp *.h /usr/local/cuda-8.0/include/

4）切换到 /usr/local/lib 目录下：sudo chmod +r libcudnn.so.5.1.10 && sudo ln -sf libcudnn.so.5.1.10 libcudnn.so.5 && sudo ln -sf libcudnn.so.5 libcudnn.so && sudo ldconfig

9、下载caffe源码、配置、编译：

=== 下载caffe源码 ===

1）cd到home/hcq/目录，下载caffe源码：sudo git clone https://github.com/BVLC/caffe.git


=== 配置相关文件 ===

1）cd到home/hcq/caffe，复制配置文件：sudo
cp Makefile.config.example Makefile.config

2）配置Makefile.config文件：sudo
gedit Makefile.config，去掉注释并修改的部分如下所示：

# cuDNN acceleration switch (uncomment to build with cuDNN).
USE_CUDNN := 1

# Uncomment if you're using OpenCV 3
OPENCV_VERSION := 3

# open for OpenBlas
BLAS := open
# (which should work)!
BLAS_INCLUDE := /opt/OpenBLAS/include
BLAS_LIB := /opt/OpenBLAS/lib

ANACONDA_HOME := /home/hcq/anaconda2
PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
$(ANACONDA_HOME)/include/python2.7 \
$(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include

PYTHON_LIB := $(ANACONDA_HOME)/lib

# Whatever else you find you need goes here.
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

# N.B. both build and distribute dirs are cleared on `make clean`
BUILD_DIR := build
DISTRIBUTE_DIR := distribute

# The ID of the GPU that 'make runtest' will use to run unit tests.
TEST_GPUID := 0

# enable pretty build (comment to see full commands)
Q ?= @

3）修改Makefile文件：sudo
gedit Makefile，进行如下修改：

NVCCFLAGS += -ccbin=$(CXX) -Xcompiler -fPIC $(COMMON_FLAGS)  # 这行去掉
NVCCFLAGS += -D_FORCE_INLINES -ccbin=$(CXX) -Xcompiler -fPIC $(COMMON_FLAGS) # 添加这行

4）进入/usr/lib/x86_64-linux-gnu目录下，执行如下命令（否则会提示：/usr/bin/ld:
找不到 -lhdf5_hl 和 /usr/bin/ld: 找不到 -lhdf5）

sudo ln libhdf5_serial.so.10.1.0 libhdf5.so
sudo ln libhdf5_serial_hl.so.10.0.2 libhdf5_hl.so
sudo ldconfig #使连接生效

5）caffe源码中，python的一些必要安装：

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

注：如果需要执行权限问题，就改变caffe文件的执行权限  sudo chown -R 'hcq' xxx

=== 编译caffe（期间遇到问题可以到第四部分查看是否有相关的解决方案）===

1）cd /home/hcq/caffe：进入caffe的根目录下

2）sudo make clean

3）sudo make all -j4 （-j4表示使用4核处理器执行当前指令）

4）sudo make test -j4（最好加上sudo防止有些文件的访问权限不够）

5）sudo make runtest -j4

6）sudo make pycaffe -j4（配置pycaffe）

=== 效果图展示 ===

1）sudo make all 成功后的效果图，如下所示：



2）sudo make runtest 成功后的效果图，如下所示：

[       OK ] ArgMaxLayerTest/1.TestCPUMaxVal (1 ms)
[ RUN      ] ArgMaxLayerTest/1.TestSetupMaxVal
[       OK ] ArgMaxLayerTest/1.TestSetupMaxVal (1 ms)
[ RUN      ] ArgMaxLayerTest/1.TestSetupAxis
[       OK ] ArgMaxLayerTest/1.TestSetupAxis (0 ms)
[ RUN      ] ArgMaxLayerTest/1.TestCPUTopK
[       OK ] ArgMaxLayerTest/1.TestCPUTopK (1 ms)
[----------] 12 tests from ArgMaxLayerTest/1 (48 ms total)

[----------] 4 tests from ContrastiveLossLayerTest/1, where TypeParam = caffe::CPUDevice<double>
[ RUN      ] ContrastiveLossLayerTest/1.TestGradientLegacy
[       OK ] ContrastiveLossLayerTest/1.TestGradientLegacy (125 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestForward
[       OK ] ContrastiveLossLayerTest/1.TestForward (0 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestGradient
[       OK ] ContrastiveLossLayerTest/1.TestGradient (136 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestForwardLegacy
[       OK ] ContrastiveLossLayerTest/1.TestForwardLegacy (0 ms)
[----------] 4 tests from ContrastiveLossLayerTest/1 (261 ms total)

[----------] Global test environment tear-down
[==========] 1106 tests from 150 test cases ran. (43894 ms total)
[  PASSED  ] 1106 tests.
hcq@hcq-To-be-filled-by-O-E-M:~/caffe$

10、caffe的[b]测试案例：[/b]

参考网址：http://caffe.berkeleyvision.org/gathered/examples/mnist.html

1）import caffe，输入如下命令：

== cd /home/hcq/caffe

== python

== import caffe

== 第一次成功后如下所示：

hcq@ubuntu:~/caffe/python$ python
Python 2.7.13 |Anaconda custom (64-bit)| (default, Dec 20 2016, 23:09:15)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
Anaconda is brought to you by Continuum Analytics.
Please check out: http://continuum.io/thanks and https://anaconda.org >>> import caffe
/home/hcq/anaconda2/lib/python2.7/site-packages/matplotlib/font_manager.py:280: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.
'Matplotlib is building the font cache using fc-list. '

2）MNIST 模型

== cd /home/hcq/caffe（确保当前目录是caffe目录），执行如下命令：

== sudo ./data/mnist/get_mnist.sh

== sudo ./examples/mnist/create_mnist.sh

== sudo ./examples/mnist/train_lenet.sh

== 训练完成后如下所示：

I0601 13:09:21.129083 29215 solver.cpp:447] Snapshotting to binary proto file examples/mnist/lenet_iter_10000.caffemodel
I0601 13:09:21.134089 29215 sgd_solver.cpp:273] Snapshotting solver state to binary proto file examples/mnist/lenet_iter_10000.solverstate
I0601 13:09:21.163990 29215 solver.cpp:310] Iteration 10000, loss = 0.00275877
I0601 13:09:21.164013 29215 solver.cpp:330] Iteration 10000, Testing net (#0)
I0601 13:09:25.226337 29217 data_layer.cpp:73] Restarting data prefetching from start.
I0601 13:09:25.394167 29215 solver.cpp:397]     Test net output #0: accuracy = 0.991
I0601 13:09:25.394209 29215 solver.cpp:397]     Test net output #1: loss = 0.0280903 (* 1 = 0.0280903 loss)
I0601 13:09:25.394215 29215 solver.cpp:315] Optimization Done.
I0601 13:09:25.394234 29215 caffe.cpp:259] Optimization Done.

3）Cifar-10 模型

== cd /home/hcq/caffe                                     #cd到caffe源码的根目录下

== sudo ./data/cifar10/get_cifar10.sh              #该脚本会下载二进制的cifar，并解压，会在/data/cifar10中出现很多batch文件 

== sudo ./examples/cifar10/create_cifar10.sh #运行后将会在examples中出现数据集./cifar10_xxx_lmdb和数据集图像均值./mean.binaryproto

注：CIFAR-10的卷积神经网络模型由卷积层，pooling层，ReLU，非线性变换层，局部对比归一化线性分类器组成。该模型定义在CAFFE_ROOT/examples/cifar10/cifar10_quick_train_test.prototxt中。

== sudo ./examples/cifar10/train_quick.sh      #先以0.001的学习率迭代4000次，再以0.01的学习率接着再迭代1000次，共5000次

注：同理可以训练full模型，full模型比quick模型迭代次数多，一共迭代70000次，前60000次学习率是0.001，中间5000次学习率是0.0001，最后5000次学习率是0.00001。full模型的网络层数也比quick模型多，输入如下命令训练full模型

== sudo ./examples/cifar10/train_full.sh

参考网址：

== caffe示例实现之1在CIFAR-10数据集上训练与测试Caffe：http://blog.csdn.net/liumaolincycle/article/details/47258937

== caffe官网：http://caffe.berkeleyvision.org/

训练成功后，提示如下所示，以sudo ./examples/cifar10/train_quick.sh为例：

I0601 17:15:20.219635 32594 solver.cpp:237]     Train net output #0: loss = 0.550304 (* 1 = 0.550304 loss)
I0601 17:15:20.219642 32594 sgd_solver.cpp:105] Iteration 4700, lr = 0.0001
I0601 17:16:15.671823 32594 solver.cpp:218] Iteration 4800 (1.80336 iter/s, 55.452s/100 iters), loss = 0.403176
I0601 17:16:15.671891 32594 solver.cpp:237]     Train net output #0: loss = 0.403176 (* 1 = 0.403176 loss)
I0601 17:16:15.671898 32594 sgd_solver.cpp:105] Iteration 4800, lr = 0.0001
I0601 17:17:10.761785 32594 solver.cpp:218] Iteration 4900 (1.81524 iter/s, 55.089s/100 iters), loss = 0.466545
I0601 17:17:10.761903 32594 solver.cpp:237]     Train net output #0: loss = 0.466545 (* 1 = 0.466545 loss)
I0601 17:17:10.761907 32594 sgd_solver.cpp:105] Iteration 4900, lr = 0.0001
I0601 17:18:02.532764 32595 data_layer.cpp:73] Restarting data prefetching from start.
I0601 17:18:04.706779 32594 solver.cpp:457] Snapshotting to HDF5 file examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5
I0601 17:18:04.708353 32594 sgd_solver.cpp:283] Snapshotting solver state to HDF5 file examples/cifar10/cifar10_quick_iter_5000.solverstate.h5
I0601 17:18:04.929492 32594 solver.cpp:310] Iteration 5000, loss = 0.451509
I0601 17:18:04.929522 32594 solver.cpp:330] Iteration 5000, Testing net (#0)
I0601 17:18:26.008786 32597 data_layer.cpp:73] Restarting data prefetching from start.
I0601 17:18:26.888097 32594 solver.cpp:397]     Test net output #0: accuracy = 0.7575
I0601 17:18:26.888123 32594 solver.cpp:397]     Test net output #1: loss = 0.737631 (* 1 = 0.737631 loss)
I0601 17:18:26.888141 32594 solver.cpp:315] Optimization Done.
I0601 17:18:26.888144 32594 caffe.cpp:259] Optimization Done.

4）使用训练的模型来分类新数据：

== 确保 sudo make pycaffe 已经编译成功。

== 打开终端，在caffe源码的根目录下，执行如下命令：

python python/classify.py examples/images/cat.jpg foo    # 直接用一下别人的模型分类试一下：（默认用的ImageNet的模型）

hcq@hcq-To-be-filled-by-O-E-M:~/caffe$ python python/classify.py examples/images/cat.jpg foo 
CPU mode
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0602 11:25:53.561646  9707 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0602 11:25:53.561698  9707 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0602 11:25:53.561709  9707 _caffe.cpp:142] Net('python/../models/bvlc_reference_caffenet/deploy.prototxt', 1, weights='python/../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel')
Traceback (most recent call last):
  File "python/classify.py", line 138, in <module>
    main(sys.argv)
  File "python/classify.py", line 110, in main
    channel_swap=channel_swap)
  File "/home/hcq/caffe/python/caffe/classifier.py", line 26, in __init__
    caffe.Net.__init__(self, model_file, pretrained_file, caffe.TEST)
RuntimeError: Could not open file python/../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel

上面这句话的意思是，使用cifar10_quick.prototxt网络 + cifar10_quick_iter_4000.caffemodel.h5模型，对examples/images/cat.jpg图片进行分类。

错误原因：

解决方案：

1）cd /home/hcq/caffe/script

2）python download_model_binary.py /home/hcq/caffe/models/bvlc_reference_caffenet

参考网址：http://blog.csdn.net/u011636440/article/details/52441519

解决上述问题后，又出现新问题提示如下所示：

I0602 11:51:24.163389 10753 upgrade_proto.cpp:53] Attempting to upgrade input file specified using deprecated V1LayerParameter: python/../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel
I0602 11:51:24.368646 10753 upgrade_proto.cpp:61] Successfully upgraded file specified using deprecated V1LayerParameter
I0602 11:51:24.436149 10753 net.cpp:744] Ignoring source layer loss
Traceback (most recent call last):
File "python/classify.py", line 138, in <module>
main(sys.argv)
File "python/classify.py", line 110, in main
channel_swap=channel_swap)
File "/home/hcq/caffe/python/caffe/classifier.py", line 34, in __init__
self.transformer.set_mean(in_, mean)
File "/home/hcq/caffe/python/caffe/io.py", line 259, in set_mean
raise ValueError('Mean shape incompatible with input shape.')
ValueError: Mean shape incompatible with input shape.

错误原因：ValueError: Mean shape incompatible with input shape. 是因为‘mean’与‘input’的尺寸不同造成的。

解决方案：修改了 ./python/caffe/io.py
文件，如下

将：

if ms != self.inputs[in_][1:]:
raise ValueError('Mean shape incompatible with input shape.')

修改为：

if ms != self.inputs[in_][1:]:
print(self.inputs[in_])
in_shape = self.inputs[in_][1:]
m_min, m_max = mean.min(), mean.max()
normal_mean = (mean - m_min) / (m_max - m_min)
mean = resize_image(normal_mean.transpose((1,2,0)),
in_shape[1:]).transpose((2,0,1)) * \
(m_max - m_min) + m_min
#raise ValueError('Mean shape incompatible with input shape.')

参考网址：http://blog.csdn.net/gzljss/article/details/45849013

运行成功后，提示如下所示：

I0602 12:00:07.864248 10953 net.cpp:200] conv1 does not need backward computation.
I0602 12:00:07.864251 10953 net.cpp:200] data does not need backward computation.
I0602 12:00:07.864254 10953 net.cpp:242] This network produces output prob
I0602 12:00:07.864266 10953 net.cpp:255] Network initialization done.
I0602 12:00:07.987157 10953 upgrade_proto.cpp:44] Attempting to upgrade input file specified using deprecated transformation parameters: python/../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel
I0602 12:00:07.987190 10953 upgrade_proto.cpp:47] Successfully upgraded file specified using deprecated data transformation parameters.
W0602 12:00:07.987195 10953 upgrade_proto.cpp:49] Note that future Caffe releases will only support transform_param messages for transformation fields.
I0602 12:00:07.987198 10953 upgrade_proto.cpp:53] Attempting to upgrade input file specified using deprecated V1LayerParameter: python/../models/bvlc_reference_caffenet/bvlc_reference_caffenet.caffemodel
I0602 12:00:08.217313 10953 upgrade_proto.cpp:61] Successfully upgraded file specified using deprecated V1LayerParameter
I0602 12:00:08.289186 10953 net.cpp:744] Ignoring source layer loss
(10, 3, 227, 227)
Loading file: examples/images/cat.jpg
Classifying 1 inputs.
Done in 0.43 s.
Saving results into foo

== 使用cifar10_quick_iter_4000.caffemodel.h5模型：

python python/classify.py --model_def examples/cifar10/cifar10_quick.prototxt --pretrained_model examples/cifar10/cifar10_quick_iter_4000.caffemodel.h5 --center_only  examples/images/cat.jpg foo # 指定自己的模型进行分类

== 错误提示如下：

hcq@hcq-To-be-filled-by-O-E-M:~/caffe$ python python/classify.py --model_def examples/cifar10/cifar10_quick.prototxt --pretrained_model examples/cifar10/cifar10_quick_iter_4000.caffemodel.h5 --center_only  examples/images/cat.jpg foo
CPU mode
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0602 11:27:22.857719  9729 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0602 11:27:22.857756  9729 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0602 11:27:22.857759  9729 _caffe.cpp:142] Net('examples/cifar10/cifar10_quick.prototxt', 1, weights='examples/cifar10/cifar10_quick_iter_4000.caffemodel.h5')
Traceback (most recent call last):
File "python/classify.py", line 138, in <module>
main(sys.argv)
File "python/classify.py", line 110, in main
channel_swap=channel_swap)
File "/home/hcq/caffe/python/caffe/classifier.py", line 26, in __init__
caffe.Net.__init__(self, model_file, pretrained_file, caffe.TEST)
RuntimeError: Could not open file examples/cifar10/cifar10_quick_iter_4000.caffemodel.h5

原因：在目录/home/hcq/caffe/examples/cifar10下，确实不存在cifar10_quick_iter_4000.caffemodel.h5这个模型，运行结果必然错误。

解决方案：到目录/home/hcq/caffe/examples/cifar10下查看已有的模型，然后把指定的模型换成已有的模型，已有的模型如下所示（因人而异）：

hcq@hcq-To-be-filled-by-O-E-M:~/caffe/examples/cifar10$ ls -lrt
总用量 6788
-rwxrwxr-x 1 hcq hcq     338 5月  27 11:55 train_quick.sh
-rwxrwxr-x 1 hcq hcq     139 5月  27 11:55 train_full_sigmoid.sh
-rwxrwxr-x 1 hcq hcq     142 5月  27 11:55 train_full_sigmoid_bn.sh
-rwxrwxr-x 1 hcq hcq     524 5月  27 11:55 train_full.sh
-rw-rw-r-- 1 hcq hcq    5243 5月  27 11:55 readme.md
-rwxrwxr-x 1 hcq hcq     467 5月  27 11:55 create_cifar10.sh
-rw-rw-r-- 1 hcq hcq    3677 5月  27 11:55 convert_cifar_data.cpp
-rw-rw-r-- 1 hcq hcq    1921 5月  27 11:55 cifar10_quick.prototxt
-rw-rw-r-- 1 hcq hcq    3122 5月  27 11:55 cifar10_full_train_test.prototxt
-rw-rw-r-- 1 hcq hcq    2879 5月  27 11:55 cifar10_full_sigmoid_train_test.prototxt
-rw-rw-r-- 1 hcq hcq    3192 5月  27 11:55 cifar10_full_sigmoid_train_test_bn.prototxt
-rw-rw-r-- 1 hcq hcq    2174 5月  27 11:55 cifar10_full.prototxt
drwxr--r-- 2 hcq root   4096 6月   1 15:38 cifar10_train_lmdb
drwxr--r-- 2 hcq root   4096 6月   1 15:38 cifar10_test_lmdb
-rw-r--r-- 1 hcq root  12299 6月   1 15:38 mean.binaryproto
-rw-rw-r-- 1 hcq hcq    3088 6月   1 15:39 cifar10_quick_train_test.prototxt
-rw-rw-r-- 1 hcq hcq     859 6月   1 16:25 cifar10_quick_solver.prototxt
-rw-rw-r-- 1 hcq hcq     944 6月   1 16:25 cifar10_full_solver_lr1.prototxt
-rw-rw-r-- 1 hcq hcq     882 6月   1 16:26 cifar10_quick_solver_lr1.prototxt
-rw-rw-r-- 1 hcq hcq     944 6月   1 16:26 cifar10_full_solver.prototxt
-rw-rw-r-- 1 hcq hcq     959 6月   1 16:26 cifar10_full_sigmoid_solver_bn.prototxt
-rw-rw-r-- 1 hcq hcq     953 6月   1 16:26 cifar10_full_sigmoid_solver.prototxt
-rw-rw-r-- 1 hcq hcq     945 6月   1 16:26 cifar10_full_solver_lr2.prototxt
-rw-r--r-- 1 hcq root 366048 6月   2 01:37 cifar10_full_iter_10000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 01:37 cifar10_full_iter_10000.caffemodel.h5
-rw-r--r-- 1 hcq root 366048 6月   2 02:34 cifar10_full_iter_20000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 02:34 cifar10_full_iter_20000.caffemodel.h5
-rw-r--r-- 1 hcq root 366048 6月   2 03:30 cifar10_full_iter_30000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 03:30 cifar10_full_iter_30000.caffemodel.h5
-rw-r--r-- 1 hcq root 366048 6月   2 04:27 cifar10_full_iter_40000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 04:27 cifar10_full_iter_40000.caffemodel.h5
-rw-r--r-- 1 hcq root 366048 6月   2 05:24 cifar10_full_iter_50000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 05:24 cifar10_full_iter_50000.caffemodel.h5
-rw-r--r-- 1 hcq root 366048 6月   2 06:21 cifar10_full_iter_60000.solverstate.h5
-rw-r--r-- 1 hcq root 375752 6月   2 06:21 cifar10_full_iter_60000.caffemodel.h5
-rw-r--r-- 1 hcq root 582496 6月   2 16:40 cifar10_quick_iter_4000.solverstate
-rw-r--r-- 1 hcq root 583301 6月   2 16:40 cifar10_quick_iter_4000.caffemodel
-rw-r--r-- 1 hcq root 590064 6月   2 16:44 cifar10_quick_iter_5000.solverstate.h5
-rw-r--r-- 1 hcq root 600032 6月   2 16:44 cifar10_quick_iter_5000.caffemodel.h5

== 选择【cifar10_quick_iter_5000.caffemodel.h5】作为指定模型，再次运行如下命令：

python python/classify.py --model_def examples/cifar10/cifar10_quick.prototxt --pretrained_model examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 --center_only  examples/images/cat.jpg foo

== 运行成功后，提示如下所示：

CPU mode
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0602 16:48:24.261683 12766 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0602 16:48:24.261732 12766 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0602 16:48:24.261735 12766 _caffe.cpp:142] Net('examples/cifar10/cifar10_quick.prototxt', 1, weights='examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5')
I0602 16:48:24.262830 12766 net.cpp:51] Initializing net from parameters:
name: "CIFAR10_quick_test"
state {
phase: TEST
level: 0
}
layer {
...
...
I0602 16:48:24.263774 12766 net.cpp:242] This network produces output prob
I0602 16:48:24.263782 12766 net.cpp:255] Network initialization done.
I0602 16:48:24.264539 12766 net.cpp:798] Ignoring source layer cifar
I0602 16:48:24.264627 12766 hdf5.cpp:32] Datatype class: H5T_FLOAT
I0602 16:48:24.265048 12766 net.cpp:798] Ignoring source layer loss
(1, 3, 32, 32)
Loading file: examples/images/cat.jpg
Classifying 1 inputs.
Done in 0.03 s.
Saving results into foo

注：默认的classify脚本不会直接输出结果，而是会把结果输入到foo文件里，不太直观，网上有一个修改版，添加了一些参数，可以输出概率最高的分类。

替换Python/classify.py，下载地址：

这个脚本添加了两个参数，可以指定labels_file，然后可以直接把分类结果输出出来：

python python/classify.py --print_results --model_def examples/cifar10/cifar10_quick.prototxt --pretrained_model examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 --labels_file data/cifar10/cifar10_words.txt  --center_only  examples/images/cat.jpg foo

Caffe学习资料：

1、Classification: Instant Recognition with Caffe：http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/00-classification.ipynb 

三、CPU版caffe环境配置过程

1、CPU和GPU版本的一些区别：

1）CPU版本的话，不需要NVIDIA驱动，不需要安装CUDA，不需要配置CUDNN

2）caffe目录下的Makefile.configure配置不一样

3）配置模型的时候，solver_mode: CPU，变成CPU（如果没改的话会报错），如下所示：



2、CPU版本的Makefile.configure配置，未注释掉的部分，如下所示：

# CPU-only switch (uncomment to build without GPU support).
CPU_ONLY := 1

# Uncomment if you're using OpenCV 3
OPENCV_VERSION := 3

# open for OpenBlas
BLAS := open
# (which should work)!
BLAS_INCLUDE := /opt/OpenBLAS/include
BLAS_LIB := /opt/OpenBLAS/lib

ANACONDA_HOME := /home/hcq/anaconda2
PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
$(ANACONDA_HOME)/include/python2.7 \
$(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include

PYTHON_LIB := $(ANACONDA_HOME)/lib

# Whatever else you find you need goes here.
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

# N.B. both build and distribute dirs are cleared on `make clean`
BUILD_DIR := build
DISTRIBUTE_DIR := distribute

# The ID of the GPU that 'make runtest' will use to run unit tests.
TEST_GPUID := 0

# enable pretty build (comment to see full commands)
Q ?= @

3、测试案例跟GPU的类似，只不过需要把solver_mode: CPU改为CPU。

四、问题集锦

1、sudo make all 编译过程遇到的错误，【conda install libgcc】安装libgcc即可，错误提示如下所示：

/usr/local/lib/libglog.so：对‘std::__cxx11::basic_ostringstream<char, std::char_traits<char>, std::allocator<char> >::~basic_ostringstream()@GLIBCXX_3.4.21’未定义的引用
/usr/lib/x86_64-linux-gnu/libgflags.so：对‘std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::find(char const*, unsigned long, unsigned long) const@GLIBCXX_3.4.21’未定义的引用
.build_release/lib/libcaffe.so：对‘std::out_of_range::out_of_range(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)’未定义的引用
.build_release/lib/libcaffe.so：对‘std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_replace_aux(unsigned long, unsigned long, unsigned long, char)’未定义的引用
/usr/local/lib/libopencv_core.so：对‘std::basic_istream<char, std::char_traits<char> >& std::getline<char, std::char_traits<char>, std::allocator<char> >(std::basic_istream<char, std::char_traits<char> >&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, char)@GLIBCXX_3.4.21’未定义的引用
.build_release/lib/libcaffe.so：对‘std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::compare(unsigned long, unsigned long, char const*) const’未定义的引用
/usr/lib/x86_64-linux-gnu/libprotobuf.so：对‘std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::swap(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&)@GLIBCXX_3.4.21’未定义的引用
/usr/local/lib/libglog.so：对‘vtable for std::__cxx11::basic_stringbuf<char, std::char_traits<char>, std::allocator<char> >@GLIBCXX_3.4.21’未定义的引用
/usr/local/lib/libopencv_core.so：对‘std::__cxx11::basic_stringbuf<char, std::char_traits<char>, std::allocator<char> >::str() const@GLIBCXX_3.4.21’未定义的引用
collect2: error: ld returned 1 exit status
Makefile:629: recipe for target '.build_release/tools/upgrade_solver_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_solver_proto_text.bin] Error 1

或者如下所示：

AR -o .build_release/lib/libcaffe.a
LD -o .build_release/lib/libcaffe.so.1.0.0
CXX/LD -o .build_release/tools/upgrade_solver_proto_text.bin
CXX/LD -o .build_release/tools/test_net.bin
CXX/LD -o .build_release/tools/upgrade_net_proto_text.bin
CXX/LD -o .build_release/tools/finetune_net.bin
.build_release/tools/test_net.o: In function `boost::system::system_error::what() const':
test_net.cpp:(.text._ZNK5boost6system12system_error4whatEv[_ZNK5boost6system12system_error4whatEv]+0x74): undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_replace(unsigned long, unsigned long, char const*, unsigned long)'
test_net.cpp:(.text._ZNK5boost6system12system_error4whatEv[_ZNK5boost6system12system_error4whatEv]+0xa1): undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_append(char const*, unsigned long)'
test_net.cpp:(.text._ZNK5boost6system12system_error4whatEv[_ZNK5boost6system12system_error4whatEv]+0xe3): undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_append(char const*, unsigned long)'
/usr/local/lib/libglog.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_create(unsigned long&, unsigned long)@GLIBCXX_3.4.21'
/usr/local/lib/libglog.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_replace_aux(unsigned long, unsigned long, unsigned long, char)@GLIBCXX_3.4.21'
...
...
...
.build_release/lib/libcaffe.so: undefined reference to `std::out_of_range::out_of_range(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_replace_aux(unsigned long, unsigned long, unsigned long, char)'
/usr/local/lib/libopencv_core.so: undefined reference to `std::basic_istream<char, std::char_traits<char> >& std::getline<char, std::char_traits<char>, std::allocator<char> >(std::basic_istream<char, std::char_traits<char> >&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&, char)@GLIBCXX_3.4.21'
.build_release/lib/libcaffe.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::compare(unsigned long, unsigned long, char const*) const'
/usr/lib/gcc/x86_64-linux-gnu/5/../../../x86_64-linux-gnu/libprotobuf.so: undefined reference to `std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::swap(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >&)@GLIBCXX_3.4.21'
/usr/local/lib/libglog.so: undefined reference to `vtable for std::__cxx11::basic_stringbuf<char, std::char_traits<char>, std::allocator<char> >@GLIBCXX_3.4.21'
/usr/local/lib/libopencv_core.so: undefined reference to `std::__cxx11::basic_stringbuf<char, std::char_traits<char>, std::allocator<char> >::str() const@GLIBCXX_3.4.21'
collect2: error: ld returned 1 exit status
Makefile:629: recipe for target '.build_release/tools/upgrade_net_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_text.bin] Error 1

2、sudo make runtest 时遇到的错误，【sudo apt install ffmpeg】安装ffmpeg即可，错误提示如下所示：

hcq@ubuntu:~/caffe$ sudo make runtest
.build_release/tools/caffe
.build_release/tools/caffe: error while loading shared libraries: libopencv_core.so.3.2: cannot open shared object file: No such file or directory
Makefile:538: recipe for target 'runtest' failed
make: *** [runtest] Error 127

3、sudo make runtest 时遇到的错误，缺失一些相关的so链接：

hcq@hcq-To-be-filled-by-O-E-M:~/caffe$ sudo make runtest -j4
.build_release/tools/caffe
.build_release/tools/caffe: error while loading shared libraries: libhdf5_hl.so.10: cannot open shared object file: No such file or directory
Makefile:536: recipe for target 'runtest' failed
make: *** [runtest] Error 127

以及

hcq@hcq-To-be-filled-by-O-E-M:~/caffe$ sudo make runtest -j4
.build_release/tools/caffe
.build_release/tools/caffe: error while loading shared libraries: libhdf5.so.10: cannot open shared object file: No such file or directory
Makefile:536: recipe for target 'runtest' failed
make: *** [runtest] Error 127

原因：目录/usr/lib/x86_64-linux-gnu下确实是找不到【libhdf5_hl.so.10】和【libhdf5.so.10】这两个文件，如下图所示：



解决方案：根据相应的文件，创建所需的软连接，如下所示：

sudo ln libhdf5_serial_hl.so.10 libhdf5_hl.so.10
sudo ln libhdf5_serial.so.10 libhdf5.so.10

4、Makefile.configure配置中，关于nvcc warning的一些警告，注释掉相关的配置代码即可，提示如下所示：

NVCC src/caffe/layers/mvn_layer.cu
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
NVCC src/caffe/layers/filter_layer.cu
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).

5、Check failed: status == CUDNN_STATUS_SUCCESS (3 vs. 0)  CUDNN_STATUS_BAD_PARAM

F0601 10:03:16.683940 28607 cudnn.hpp:122] Check failed: status == CUDNN_STATUS_SUCCESS (3 vs. 0)  CUDNN_STATUS_BAD_PARAM
*** Check failure stack trace: ***
@     0x7f52e7de2e3d  google::LogMessage::Fail()
@     0x7f52e7de4bc0  google::LogMessage::SendToLog()
@     0x7f52e7de2a23  google::LogMessage::Flush()
@     0x7f52e7de558e  google::LogMessageFatal::~LogMessageFatal()
@     0x7f52e313c5f8  caffe::CuDNNConvolutionLayer<>::Reshape()
@     0x7f52e325ed37  caffe::Net<>::Init()
@     0x7f52e326146e  caffe::Net<>::Net()
@           0x888f2c  caffe::NetTest<>::InitNetFromProtoString()
@           0x88e1d1  caffe::NetTest_TestReshape_Test<>::TestBody()
@           0x91e4b3  testing::internal::HandleExceptionsInMethodIfSupported<>()
@           0x917aca  testing::Test::Run()
@           0x917c18  testing::TestInfo::Run()
@           0x917cf5  testing::TestCase::Run()
@           0x918fcf  testing::internal::UnitTestImpl::RunAllTests()
@           0x9192f3  testing::UnitTest::Run()
@           0x46d44d  main
@     0x7f52e2443830  (unknown)
@           0x474f69  _start
Makefile:536: recipe for target 'runtest' failed
make: *** [runtest] 已放弃 (core dumped)

原因：初步断定是NVIDIA的显存不够，本文所使用的机器是GeForce GTX 960  2GB的显存。本人做了如下探究：

1）据同学介绍，他使用的机器显存有8GB，就不会出现这样的问题。

2）本人还尝试使用CPU( 物理内存8GB )运行sudo make runtest -j4，也不会出现这样的问题。

解决方案：

1）从机器层面入手，扩大硬件性能（主要是显存）

2）改用CPU跑程序，不过最后使用GPU的话还是需要使用好一点的设备（显卡、显卡）

参考网址：http://www.cnblogs.com/penguinliong/p/6351215.html

6、build_release/tools/caffe: error while loading shared libraries: libopenblas.so.0: cannot open shared object file: No such file or directory

描述：Makefile.configure配置好了，相应的引用位置也有 libopenblas.so.0 该文件，make runtest的时候仍然出错。

解决方案：将 openblas 的库所在位置添加到系统环境变量 LD_LIBRARY_PATH，export LD_LIBRARY_PATH=/opt/OpenBLAS/lib/ sudo ldconfig。打开终端，输入如下命令：

1）export LD_LIBRARY_PATH=/opt/OpenBLAS/lib/ 

2）sudo ldconfig

参考网址：

1）http://blog.csdn.net/lanxuecc/article/details/51997919

2）http://blog.csdn.net/lien0906/article/details/46816243

3）https://github.com/sermanet/OverFeat/issues/10

配置成功后的提示如下所示：

[       OK ] ArgMaxLayerTest/1.TestCPUMaxVal (1 ms)
[ RUN      ] ArgMaxLayerTest/1.TestSetupMaxVal
[       OK ] ArgMaxLayerTest/1.TestSetupMaxVal (1 ms)
[ RUN      ] ArgMaxLayerTest/1.TestSetupAxis
[       OK ] ArgMaxLayerTest/1.TestSetupAxis (0 ms)
[ RUN      ] ArgMaxLayerTest/1.TestCPUTopK
[       OK ] ArgMaxLayerTest/1.TestCPUTopK (1 ms)
[----------] 12 tests from ArgMaxLayerTest/1 (48 ms total)

[----------] 4 tests from ContrastiveLossLayerTest/1, where TypeParam = caffe::CPUDevice<double>
[ RUN      ] ContrastiveLossLayerTest/1.TestGradientLegacy
[       OK ] ContrastiveLossLayerTest/1.TestGradientLegacy (125 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestForward
[       OK ] ContrastiveLossLayerTest/1.TestForward (0 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestGradient
[       OK ] ContrastiveLossLayerTest/1.TestGradient (136 ms)
[ RUN      ] ContrastiveLossLayerTest/1.TestForwardLegacy
[       OK ] ContrastiveLossLayerTest/1.TestForwardLegacy (0 ms)
[----------] 4 tests from ContrastiveLossLayerTest/1 (261 ms total)

[----------] Global test environment tear-down
[==========] 1106 tests from 150 test cases ran. (43894 ms total)
[  PASSED  ] 1106 tests.
hcq@hcq-To-be-filled-by-O-E-M:~/caffe$

7、Check failed: fd != -1 (-1 vs. -1) File not found: examples/cifar10/cifar10_full_iter_60000.solverstate

描述：执行【sudo ./examples/cifar10/train_full.sh】报错。

解决方案：

相似问题：Check failed: fd != -1 (-1 vs. -1) File not found: deploy.prototxt

原因：因为没有把deploy.prototxt文件复制到demo下。

解决方案：把caffe/models/bvlc_reference_caffnnet/deploy.prototxt复制到demo下即可。

参考网址：http://blog.csdn.net/ahbbshenfeng/article/details/52077605

8、error while loading shared libraries: libhdf5_hl.so.100

build_release/tools/caffe: error while loading shared libraries: libhdf5_hl.so.100: cannot open shared object file: No such file or directory
Makefile:470: recipe for target 'runtest' failed

参考网址：http://blog.csdn.net/a8039974/article/details/77179330

9、Makefile:567: recipe for target '.build_release/examples/cifar10/convert_cifar_data.bin' failed

Makefile:567: recipe for target '.build_release/examples/mnist/convert_mnist_data.bin' failed
make: *** [.build_release/examples/mnist/convert_mnist_data.bin] Error 1
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadRGBAStrip@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFIsTiled@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFWriteScanline@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFGetField@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFScanlineSize@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadEncodedTile@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadRGBATile@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFClose@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFRGBAImageOK@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFOpen@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadEncodedStrip@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetField@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetWarningHandler@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetErrorHandler@LIBTIFF_4.0'
collect2: error: ld returned 1 exit status
Makefile:567: recipe for target '.build_release/examples/cifar10/convert_cifar_data.bin' failed
make: *** [.build_release/examples/cifar10/convert_cifar_data.bin] Error 1

reference：https://www.cnblogs.com/whu-zeng/p/6874307.html

9、Ubuntu 16.04中使用Matlab R2016b编译Caffe接口出错的解决方法

MEX-file 'caffe/matlab/+caffe/private/caffe_.mexa64' 无效:
caffe/matlab/+caffe/private/caffe_.mexa64: undefined
symbol: _ZN2cv8imencodeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEERKNS_11_InputArrayERSt6vectorIhSaIhEERKSB_IiSaIiEE。

出错 caffe.set_mode_cpu (line 5)
caffe_('set_mode_cpu');

出错 caffe.run_tests (line 6)
caffe.set_mode_cpu();



reference：https://github.com/BVLC/caffe/issues/3934

10、gcc 降级后的问题

/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFGetField@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFScanlineSize@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::NameOfEnum(google::protobuf::EnumDescriptor const*, int)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::string const&, int)'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadEncodedTile@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadRGBATile@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFClose@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::DescriptorPool::FindFileByName(std::string const&) const'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFRGBAImageOK@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::WireFormatLite::ReadBytes(google::protobuf::io::CodedInputStream*, std::string*)'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::StringTypeHandlerBase::Delete(std::string*)'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFOpen@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::MessageFactory::InternalRegisterGeneratedFile(char const*, void (*)(std::string const&))'
.build_release/lib/libcaffe.so: undefined reference to `leveldb::DB::Open(leveldb::Options const&, std::string const&, leveldb::DB**)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std::string const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFReadEncodedStrip@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetField@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::StringTypeHandlerBase::New()'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::WireFormatLite::WriteBytesMaybeAliased(int, std::string const&, google::protobuf::io::CodedOutputStream*)'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetWarningHandler@LIBTIFF_4.0'
/usr/lib/x86_64-linux-gnu/libopencv_highgui.so: undefined reference to `TIFFSetErrorHandler@LIBTIFF_4.0'
.build_release/lib/libcaffe.so: undefined reference to `leveldb::Status::ToString() const'
.build_release/lib/libcaffe.so: undefined reference to `google::protobuf::internal::WireFormatLite::WriteString(int, std::string const&, google::protobuf::io::CodedOutputStream*)'
collect2: error: ld returned 1 exit status
Makefile:567: recipe for target '.build_release/examples/cifar10/convert_cifar_data.bin' failed
make: *** [.build_release/examples/cifar10/convert_cifar_data.bin] Error 1

reference：http://blog.csdn.net/u010734277/article/details/70195043

其他相关的参考网址：

1、解决/usr/lib/libstdc++.so.6: version `GLIBCXX_3.4.14' not found问题：http://blog.csdn.net/xiaolong2w/article/details/23915171

2、Caffe搭建：常见问题解决办法和ubuntu使用中遇到问题（持续更新）：http://www.cnblogs.com/empty16/p/4828476.html

3、安装python caffe过程中遇到的一些问题以及对应的解决方案：http://www.cnblogs.com/TiBAi/p/6848307.html

4、安装caffe总结：http://blog.csdn.net/csoldiers/article/details/51685179

5、caffe+cuda7.0+opencv3.0.0+mkl ubuntu14.04配置：http://www.th7.cn/system/lin/201504/103304.shtml

6、Caffe
环境搭建中应注意的问题：http://blog.csdn.net/xiaxiazls/article/details/52039473

7、caffe
cpu版 Anaconda3 python 接口安装：http://blog.csdn.net/u012675539/article/details/51351553