Faster R-CNN安装笔记，只用CPU

转自：http://blog.sina.com.cn/s/blog_679f93560102wpyf.html

• 下载代码和数据

git clone --recursive  https://github.com/rbgirshick/py-faster-rcnn.git

• 下载demo模型数据

[root@localhost py-faster-rcnn]#./data/scripts/fetch_faster_rcnn_models.sh

Downloading Faster R-CNN demo models (695M)...

。。。

Unzipping...

faster_rcnn_models/

faster_rcnn_models/ZF_faster_rcnn_final.caffemodel

faster_rcnn_models/VGG16_faster_rcnn_final.caffemodel

• 编译cython

进入lib目录，修改setup.py，注释掉GPU相关代码，如下

。。。

#CUDA =locate_cuda()

。。。

#           self.set_executable('compiler_so',CUDA['nvcc'])

。。。

#   Extension('nms.gpu_nms',

#       ['nms/nms_kernel.cu', 'nms/gpu_nms.pyx'],

#       library_dirs=[CUDA['lib64']],

#       libraries=['cudart'],

#       language='c++',

#       runtime_library_dirs=[CUDA['lib64']],

#       # this syntax is specific to this build system

#       # we're only going to use certain compiler args with nvcc and notwith

#       # gcc the implementation of this trick is in customize_compiler()below

#       extra_compile_args={'gcc':["-Wno-unused-function"],

#                           'nvcc': ['-arch=sm_35',

#                                    '--ptxas-options=-v',

#                                    '-c',

#                                    '--compiler-options',

#                                    "'-fPIC'"]},

#       include_dirs = [numpy_include, CUDA['include']]

#   ),

。。。

编译：

[root@localhost lib]# make

• 安装caffe（自带的，不是通用的）

进入caffe-fast-rcnn目录，大部分跟前面caffe安装记录一文一样，修改Makefile.config为

## Refer tohttp://caffe.berkeleyvision.org/installation.html

# Contributionssimplifying and improving our build system arewelcome!

# cuDNN accelerationswitch (uncomment to build with cuDNN).

# USE_CUDNN :=1

# CPU-only switch(uncomment to build without GPU support).

CPU_ONLY :=1

# uncomment to disableIO dependencies and corresponding data layers

# USE_OPENCV :=0

# USE_LEVELDB :=0

# USE_LMDB :=0

# uncomment to allowMDB_NOLOCK when reading LMDB files (only if necessary)

#   You should not set this flag if you will bereading LMDBs with any

#   possibility of simultaneous read andwrite

# ALLOW_LMDB_NOLOCK :=1

# Uncomment if you'reusing OpenCV 3

# OPENCV_VERSION :=3

# To customize yourchoice of compiler, uncomment and set the following.

# N.B. the default forLinux is g++ and the default for OSX is clang++

# CUSTOM_CXX :=g++

# CUDA directorycontains bin/ and lib/ directories that we need.

# CUDA_DIR :=/usr/local/cuda

# On Ubuntu 14.04, ifcuda tools are installed via

# "sudo apt-get installnvidia-cuda-toolkit" then use this instead:

# CUDA_DIR :=/usr

# CUDA architecturesetting: going with all of them.

# For CUDA < 6.0,comment the *_50 lines for compatibility.

#CUDA_ARCH := -gencodearch=compute_20,code=sm_20 \

#       -gencodearch=compute_20,code=sm_21 \

#       -gencodearch=compute_30,code=sm_30 \

#       -gencodearch=compute_35,code=sm_35 \

#       -gencodearch=compute_50,code=sm_50 \

#       -gencodearch=compute_50,code=compute_50

# BLASchoice:

# atlas for ATLAS(default)

# mkl forMKL

# open forOpenBlas

BLAS :=atlas

# Custom(MKL/ATLAS/OpenBLAS) include and lib directories.

# Leave commented toaccept the defaults for your choice of BLAS

# (which shouldwork)!

BLAS_INCLUDE :=/usr/include/atlas-x86_64-base

BLAS_LIB :=/usr/lib64/atlas

# Homebrew putsopenblas in a directory that is not on the standard searchpath

# BLAS_INCLUDE :=$(shell brew --prefix openblas)/include

# BLAS_LIB := $(shellbrew --prefix openblas)/lib

# This is required onlyif you will compile the matlab interface.

# MATLAB directoryshould contain the mex binary in /bin.

# MATLAB_DIR :=/usr/local

# MATLAB_DIR :=/Applications/MATLAB_R2012b.app

# NOTE: this isrequired only if you will compile the pythoninterface.

# We need to be able tofind Python.h and numpy/arrayobject.h.

PYTHON_INCLUDE :=/usr/include/python2.7 \

              /usr/lib64/python2.7/site-packages/numpy/core/include

# Anaconda Pythondistribution is quite popular. Include path:

# Verify anacondalocation, sometimes it's in root.

# ANACONDA_HOME :=$(HOME)/anaconda

# PYTHON_INCLUDE :=$(ANACONDA_HOME)/include \

       #$(ANACONDA_HOME)/include/python2.7 \

       #$(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include\

# Uncomment to usePython 3 (default is Python 2)

# PYTHON_LIBRARIES :=boost_python3 python3.5m

# PYTHON_INCLUDE :=/usr/include/python3.5m \

#             /usr/lib/python3.5/dist-packages/numpy/core/include

# We need to be able tofind libpythonX.X.so or .dylib.

PYTHON_LIB :=/usr/lib64

# PYTHON_LIB :=$(ANACONDA_HOME)/lib

# Homebrew installsnumpy in a non standard path (keg only)

# PYTHON_INCLUDE +=$(dir $(shell python -c 'import numpy.core;print(numpy.core.__file__)'))/include

# PYTHON_LIB += $(shellbrew --prefix numpy)/lib

# Uncomment to supportlayers written in Python (will link against Pythonlibs)

WITH_PYTHON_LAYER :=1

# Whatever else youfind you need goes here.

INCLUDE_DIRS :=$(PYTHON_INCLUDE) /usr/include

LIBRARY_DIRS :=$(PYTHON_LIB) /usr/lib64

# If Homebrew isinstalled at a non standard location (for example your homedirectory) and you use it for general dependencies

# INCLUDE_DIRS +=$(shell brew --prefix)/include

# LIBRARY_DIRS +=$(shell brew --prefix)/lib

# Uncomment to use`pkg-config` to specify OpenCV library paths.

# (Usually notnecessary -- OpenCV libraries are normally installed in one of theabove $LIBRARY_DIRS.)

# USE_PKG_CONFIG :=1

BUILD_DIR :=build

DISTRIBUTE_DIR :=distribute

# Uncomment fordebugging. Does not work on OSX due tohttps://github.com/BVLC/caffe/issues/171

# DEBUG :=1

# The ID of the GPUthat 'make runtest' will use to run unit tests.

# TEST_GPUID :=0

# enable pretty build(comment to see full commands)

Q ?= @

修改Makefile

LIBRARIES += satlas tatlas #新版atlas已经不用这两个lib了：cblasatlas

编译caffe和pycaffe

 [root@localhost caffe-fast-rcnn]# make -j8&& make pycaffe

• 跑demo

[root@localhost py-faster-rcnn]# ./tools/demo.py

Traceback (most recent call last):

File "./tools/demo.py", line 17, in

 from fast_rcnn.config import cfg

 File"/root/zhanxiang/work/py-faster-rcnn/tools/../lib/fast_rcnn/config.py",line 23, in

from easydict import EasyDict as edict

ImportError: No module named easydict

缺少Python库easydict，所以安装 pip install easydict

[root@localhost py-faster-rcnn]# ./tools/demo.py

Traceback (most recent call last):

  File "./tools/demo.py", line 18, in

    from fast_rcnn.testimport im_detect

  File"/root/zhanxiang/work/py-faster-rcnn/tools/../lib/fast_rcnn/test.py",line 15, in

    import cv2

ImportError: No module named cv2

缺少Python库cv2，这个是openCV里面的。那就来 装openCV python库

yum install opencv-python.x86_64

[root@localhost py-faster-rcnn]# python tools/demo.py--cpu

Traceback (most recent call last):

  File "tools/demo.py", line 21, in

    importmatplotlib.pyplot as plt

  File"/usr/lib64/python2.7/site-packages/matplotlib/pyplot.py", line 26,in

    from matplotlib.figureimport Figure, figaspect

  File"/usr/lib64/python2.7/site-packages/matplotlib/figure.py", line 36,in

    from matplotlib.axesimport Axes, SubplotBase, subplot_class_factory

  File"/usr/lib64/python2.7/site-packages/matplotlib/axes/__init__.py",line 4, in

    from ._subplots import*

  File"/usr/lib64/python2.7/site-packages/matplotlib/axes/_subplots.py",line 10, in

    frommatplotlib.axes._axes import Axes

  File"/usr/lib64/python2.7/site-packages/matplotlib/axes/_axes.py", line14, in

    from matplotlib importunpack_labeled_data

ImportError: cannot import nameunpack_labeled_data

看起来跟 matplotlib库有关，pipinstall的版本太旧， 直接下载源码安装。

按照官网指示， http://matplotlib.org/faq/installing_faq.html#install-from-git

[root@localhost work]# git clonegit://github.com/matplotlib/matplotlib.git

[root@localhost work]# cd matplotlib/

安装依赖包

[root@localhost matplotlib]# yum-builddeppython-matplotlib

安装

[root@localhost matplotlib]# python setup.py install

[root@localhost py-faster-rcnn]# python tools/demo.py--cpu

Traceback (most recent call last):

  File "tools/demo.py", line 19, in

    fromfast_rcnn.nms_wrapper import nms

  File"/root/zhanxiang/work/py-faster-rcnn/tools/../lib/fast_rcnn/nms_wrapper.py",line 9, in

    from nms.gpu_nms importgpu_nms

ImportError: No module named gpu_nms

修改nms_wrapper.py，改 force_cpu = True

[root@localhost py-faster-rcnn]# vilib/fast_rcnn/nms_wrapper.py

def nms ( dets , thresh , force_cpu = True ):

• 大功告成

[root@localhost py-faster-rcnn]# pythontools/demo.py --cpu

就能看到结果了