ubuntu使用cmake编译coffe

 教程

Caffe Installation Tips - OpenVINO™ Toolkit

1.克隆 caffe

2. (1)mkdir build

    (2)cmake ..

然后更改build/cMakeCache.txt中的CMAKE_CXX_FLAGS:STRING=-fPIC

(3)make

报错情况/usr/lib/ligflags.a

那么使用下面的方法解决

链接relocation R_X86_64_32S against `.rodata' can not be used when making a shared object; recompile with -fPIC · Issue #2171 · BVLC/caffe · GitHub

I had a similar error message and the suggested fixes above didn't work. 
But I found that I had an old installation of gflags that I had built from source, 
so I removed it by deleting the files /usr/local/lib/libgflags_nothreads.
a and /usr/local/lib/libgflags.a, in addition to sudo rm -rf /usr/local/lib/cmake/gflags and sudo rm -rf /usr/local/include/gflags/.

如果是缺少pycoffe

使用sudo apt-get install python-dev安装 

编译完成之后，如何用python来调用，参考如下

1. 使用make pycaffe -j8 来编译pyhton接口

2.使用下面的命令，来迁移包

Caffe Installation Tips - OpenVINO™ Toolkit

注意 复制caffe的时候是复制build的上级目录中的python/caffe，而不是build里面的，然后也不用复制_caffe.o这个文件，就能 直接在python中引用

3.测试

import caffe,如果报错，查看原因，原来是numpy版本有问题 ，意思是，编译的时候用的numy 16,当前的环境是numpy15,所以使用

pip install numpy --upgrade

numpy 1.23.5可以import caffe

Successfully uninstalled numpy-1.23.5

python face_seg.py 测试结果