1. 安装环境
- Ubuntu18.04.6 LTS
- Python 3.6.13
- Nvidia Geforce 1070
- CUDA 10.1
- CuDNN 8.1.1
- torch==1.8.2
2. 安装步骤
2.1 安装部分依赖文件
# python3 modules (numpy, protobuf, skimage)
sudo pip3 install numpy
sudo apt-get install python3-skimage
sudo apt-get install python3-protobuf
# build essential
sudo apt-get install build-essential cmake git pkg-config
# gflags, glog, lmdb
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev
# boost
sudo apt-get install libboost-all-dev
# hdf5
sudo apt-get install libhdf5-dev
# protobuf
sudo apt-get install protobuf-compiler libprotobuf-dev
# blas
sudo apt-get install libblas-dev libcblas-dev libatlas-base-dev libopenblas-dev
# leveldb
sudo apt-get install libleveldb-dev
# snappy
sudo apt-get install libsnappy-dev
2.2 安装OpenCV
这里以OpenCV3.3.1为例,更高版本也能正常运行
2.2.1 安装部分依赖
sudo apt-get install build-essential
sudo apt-get install cmake git libgtk2.0-dev pkg-config libavcodec-dev libavformat-dev libswscale-dev
sudo apt-get install python-dev python-numpy libtbb2 libtbb-dev libjpeg-dev libpng-dev libtiff-dev libjasper-dev libdc1394-22-dev
2.2.2 下载源码
下载地址:https://github.com/opencv/opencv/archive/refs/tags/3.3.1.tar.gz
下载后解压到安装目录下,解压后的文件名为opencv-3.3.1。
2.2.3 开始安装
2.2.3.1 预备阶段
cd "Your Path"/opencv-3.3.1
mkdir build # 创建工程编译所需文件夹
cd build
2.2.3.2 配置文件
CMAKE_BUILD_TYPE =RELEASE指明编译的版本是Release版,CMAKE_INSTALL_PREFIX=/usr/local指明编译后的可执行程序的存放目录。
注意,后面的两个点千万不能省,代表了上级目录
cmake -D CMAKE_BUILD_TYPE=Release -D CMAKE_INSTALL_PREFIX=/usr/local ..
2.2.3.3 编译和安装
-j8指8线程同时编译,根据自己CPU的特点调整该参数,如果不了解自己CPU,直接执行make即可。
make -j8 # 多线程执行make任务
sudo make install
2.2.3.4 查看安装信息
pkg-config --modversion opencv #查看安装版本
2.2.3.5 验证程序
创建main.cpp文件,加入如下代码:
#include <iostream>
#include <opencv2/opencv.hpp>
using namespace std;
using namespace cv;
int main()
{
Mat srcImage = imread("lena.jpg");
imshow("Image",srcImage);
waitKey(0);
return 0;
}
2.3 安装Caffe
2.3.1 下载源码
cd ~/"Your Caffe setup route" # 开发环境目录
git clone https://github.com/BVLC/caffe
cd caffe
cp Makefile.config.example Makefile.config
2.3.2 修改Makefile.config文件
- 取消对CPU_ONLY := 1的注释(第8行)
# CPU-only switch (uncomment to build without GPU support).
CPU_ONLY := 1
- 取消对OPENCV_VERSION := 3的注释(第22行)
# Uncomment if you're using OpenCV 3
OPENCV_VERSION := 3
- 取消对PYTHON_INCLUDE的注释(第80行),并对之前的python2对应的PYTHON_INCLUDE进行注释
# NOTE: this is required only if you will compile the python interface.
# We need to be able to find Python.h and numpy/arrayobject.h.
# PYTHON_INCLUDE := /usr/include/python2.7 \
/usr/lib/python2.7/dist-packages/numpy/core/include
# Anaconda Python distribution is quite popular. Include path:
# Verify anaconda location, 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 use Python 3 (default is Python 2)
PYTHON_LIBRARIES := boost_python-py36
PYTHON_INCLUDE := /usr/include/python3.6m \
/usr/lib/python3.6/dist-packages/numpy/core/include
- 配置引用文件路径(第96行)(主要是HDF5的路径问题)
# Whatever else you find you need goes here.
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial /opt/Opencv2/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu /usr/lib/x86_64-linux-gnu/hdf5/serial /opt/Opencv2/lib
- 编译并测试
sudo make all
sudo make test
sudo make runtest
2.4 安装Caffe-Python接口
2.4.1 安装Python接口依赖库
在caffe根目录下,有个python文件夹,文件夹里面有个requirements.txt,里面有需要的依赖库和版本信息,按照其安装即可,在安装前,需要先安装fortran编辑器(gfrotran),因为安装scipy库时需要它,命令如下:
sudo apt-get install gfortran
cd "Your route"/caffe/python
for req in $(cat requirements.txt); do pip install $req; done
2.4.2 将caffe根目录下的python文件夹加入到环境变量
在caffe根目录下,有个python文件夹,文件夹里面有个requirements.txt,里面有需要的依赖库和版本信息,按照其安装即可,在安装前,需要先安装fortran编辑器(gfrotran),因为安装scipy库时需要它,命令如下:
sudo apt-get install gfortran
cd "Your route"/caffe/python
for req in $(cat requirements.txt); do pip install $req; done
2.4.2 将caffe根目录下的python文件夹加入到环境变量
2.4.2.1 先打开配置文件bashrc
sudo gedit ~/.bashrc
2.4.2.2在文件的最后面添加
export PYTHONPATH="Your Own Path"/caffe/python:$PYTHONPATH
注意,这里路径根据每个人电脑则不一样,可以自行进入自己的caffe文件夹里的python目录下,输入pwd查看路径。
2.4.2.3 然后执行更新配置(建议重启电脑)
sudo ldconfig
2.4.2.4 编译python接口
cd "Your Own Path"/caffe/
make pycaffe
2.4.2.5 验证python接口
Python 3.6.13 |Anaconda, Inc.| (default, Jun 4 2021, 14:25:59)
[GCC 7.5.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import caffe
>>>
到此Caffe全部安装完成
3. 可能遇到的问题
问题1
编译Caffe的时候出现问题"ld cannot find lboost_python3"
这个时候应该创建一个libboost_python-py36.so的符号链接。
具体做法为"cannot find -lboost_python3" when using Python3 Ubuntu16.04:
cd /usr/lib/x86_64-linux-gnu
sudo ln -s libboost_python-py36.so libboost_python3.so
问题2
在编译make test的时候显示:"libopencv_core.so.3.3: cannot open shared object: …"
修改在/etc/ld.so.conf为:
include /etc/ld.so.conf.d/*.conf
include /usr/local/lib
问题3
安装caffe遇到 Makefile:635: recipe for target '.build_release/tools/compute_image_mean.bin' failed
将Makefile.config修改为(108行):
# Uncomment to use `pkg-config` to specify OpenCV library paths.
# (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)
USE_PKG_CONFIG :=
LIBRARIES += glog gflags protobuf leveldb snappy \
lmdb boost_system hdf5_hl hdf5 m \
opencv_core opencv_highgui opencv_imgproc opencv_imgcodecs
问题4
安装Caffe遇到 undefined symbol: _ZN5boost6python6detail11init_moduleER11PyModuleDefPFvvE
原因是boost_python的版本不匹配
先在系统的/usr/lib/x86_64-linux-gnu/路径下查找boost_python文件的版本是否与python版本匹配,我的电脑中文件名为:libboost_python3-py36.so.1.65.1 说明版本为python3.6,匹配。若不匹配则安装对应的boost_python版本。
然后修改Makefile.config(80行)
# Uncomment to use Python 3 (default is Python 2)
PYTHON_LIBRARIES := boost_python-py36
接着重新编译文件!