211服务器上安装部署OpenPose

OpenPose是用来做人体检测，描绘出人体的外形关节。发布在[GitHub](https://github.com/CMU-Perceptual-Computing-Lab/openpose)上。历经一个多礼拜，我们把它安装部署在211服务器上。过程曲折，记录留念。
 OpenPose具体安装教程参考[installation](https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/doc/installation.md)，它依赖的环境如下：
 - CUDA 8.0
 - cuDNN 5.1
 - OpenCV 2.4
 - caffe 1.0
 我们系统的环境是Ubuntu 16.04，显卡GeForce GTX TITAN X
 参考[Ubuntu16.04+CUDA8.0+cuDNN5.1+caffe安装教程](https://blog.csdn.net/autocyz/article/details/52299889)
 首先安装CUDA-8.0。CUDA是由NVIDIA公司推出的一种并行运算架构，该架构能够使GPU解决复杂运算。安装CUDA之前我们需要安装NVIDIA驱动。这一点可以在[NVIDIA](https://www.nvidia.com/Download/index.aspx?lang=en-us)上根据设备的配置下载安装相应的驱动。也可以在安装CUDA时会提示是否安装drivier选择Yes。可通过命令`nvidia-smi`查看驱动。最后把CUDA的路径加入PATH。
 cuDNN相当于是CUDA的一个加速器。根据CUDA版本选择相应版本的cuDNN。比如CUDA-8.0对应cuDNN的版本是cuDNN5.1。这里需要注意的是在/usr/local目录下会有两个目录，cuda和cuda-8.0。在安装cuDNN时，我们把解压得到的文件复制到cuda相应目录下。然后建立软连接。
 最后在安装完成后，通过`nvcc--version`查看版本信息。也可以在/usr/local/cuda-8.0/samples/1_Utilities/deviceQuery目录下`make`以后运行`./deviceQuery`查看CUDA信息，若无报错，则安装成功。
 以上是个人安装CUDA+cuDNN的经历。然而实际在openpose项目中， 
 **只需要运行/openpose/ubuntu/install_cmake.sh即可！**
 接下来，按照提示`apt-get install libopencv-dev`。caffe可以选择不安装。（因为安装caffe的坑特别多！）
 然后麻烦来了，根据提示，我们在`cmake..`时，报错提示找不到cuDNN!!!这时，有两种方法解决。第一种：通过`cmake -DCUDNN_INCLUDE_DIR=/usr/local/cuda/include -DCUDNN_LIBRARY=/usr/local/cuda/lib64/libcudnn.so ..`编译。
第二种：`vim /openpose/cmake/Modules/Find_CuDNN.cmake'加入
```
set(CUDNN_INCLUDE "/usr/local/cuda/include")
set(CUDNN_LIBRARY "/usr/local/cuda/lib64/libcudnn.so")
```
 这个时候应该可以通过了，接下来在`make -j40`时，也会出现，cuDNN not find然后整个就报错。查看报错日志，报错原因是‘pthread_create’这个函数。然而我们折腾了许久找不到与这个函数相关的错误。实际上我们是被**误导**了！！最后我们把问题回到cuDNN not find上。查看寻找cudnn的路径。运行`grep 'cuda-8.0 -r ./*'`找到所有编译文件。把包含CUDNN的路径中'cuda-8.0'改为'cuda'此时即可通过。在make过程中要等一些时间，请耐心。
 在编译通过后，即可运行测试用例`./build/examples/openpose/openpose.bin --image_dir examples/media `此时会出现'CUDA driver version is insufficient for CUDA runtime version'的错误！翻译过来是CUDA的驱动版本不支持它的运行版本！也就是说我们安装的CUDA的版本不对！前面提到过，我们用`./deviceQuery`查询CUDA时，此时显示的是驱动版本CUDA 9.0, 运行版本CUDA 8.0。会不会是这个原因昵。我们尝试卸载显卡驱动（不影响CUDA）nvidia 384, 安装nvidia 375之后。使得CUDA驱动版本和运行版本一致。再次重新安装openpose。然而，报同样的错误！！！再次查找原因，发现Nvidia 375不足以支撑我们的显卡配置。在这里，索性从头再来。恢复nvidia 384。按照教程重新安装openpose。按照之前的经验，一一排除坑。发现cudnn的版本升到了6.0。果然在make报错：
```
CMakeFiles/caffe.bin.dir/caffe.cpp.o: In function `train()':
caffe.cpp:(.text+0x28ea): undefined reference to `google::protobuf::internal::StringTypeHandlerBase::New()'
caffe.cpp:(.text+0x36a8): undefined reference to `google::protobuf::internal::StringTypeHandlerBase::New()'
CMakeFiles/caffe.bin.dir/caffe.cpp.o: In function `std::string* google::MakeCheckOpString<cudaError, cudaError>(cudaError const&, cudaError const&, char const*)':
caffe.cpp:(.text._ZN6google17MakeCheckOpStringI9cudaErrorS1_EEPSsRKT_RKT0_PKc[_ZN6google17MakeCheckOpStringI9cudaErrorS1_EEPSsRKT_RKT0_PKc]+0x50): undefined reference to `google::base::CheckOpMessageBuilder::NewString()'
CMakeFiles/caffe.bin.dir/caffe.cpp.o: In function `std::string* google::MakeCheckOpString<unsigned long, int>(unsigned long const&, int const&, char const*)':
caffe.cpp:(.text._ZN6google17MakeCheckOpStringImiEEPSsRKT_RKT0_PKc[_ZN6google17MakeCheckOpStringImiEEPSsRKT_RKT0_PKc]+0x50): undefined reference to `google::base::CheckOpMessageBuilder::NewString()'
CMakeFiles/caffe.bin.dir/caffe.cpp.o: In function `main':
caffe.cpp:(.text.startup+0x50): undefined reference to `google::SetVersionString(std::string const&)'
caffe.cpp:(.text.startup+0x83): undefined reference to `google::SetUsageMessage(std::string const&)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::WireFormatLite::WriteStringMaybeAliased(int, std::string const&, google::protobuf::io::CodedOutputStream*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::io::CodedOutputStream::WriteStringWithSizeToArray(std::string const&, unsigned char*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::WireFormatLite::ReadString(google::protobuf::io::CodedInputStream*, std::string*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::Message::GetTypeName() const'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::StringTypeHandlerBase::Delete(std::string*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::MessageFactory::InternalRegisterGeneratedFile(char const*, void (*)(std::string const&))'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::Message::DebugString() const'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::WireFormatLite::WriteBytesMaybeAliased(int, std::string const&, google::protobuf::io::CodedOutputStream*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::MessageLite::ParseFromString(std::string const&)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::empty_string_'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::NameOfEnum(google::protobuf::EnumDescriptor const*, int)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::WireFormatLite::WriteString(int, std::string const&, google::protobuf::io::CodedOutputStream*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::DescriptorPool::FindFileByName(std::string const&) const'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::internal::WireFormatLite::ReadBytes(google::protobuf::io::CodedInputStream*, std::string*)'
../lib/libcaffe.so.1.0.0: undefined reference to `google::protobuf::Message::InitializationErrorString() const'
collect2: error: ld returned 1 exit status
tools/CMakeFiles/caffe.bin.dir/build.make:128: recipe for target 'tools/caffe' failed
make[5]: *** [tools/caffe] Error 1
CMakeFiles/Makefile2:425: recipe for target 'tools/CMakeFiles/caffe.bin.dir/all' failed
make[4]: *** [tools/CMakeFiles/caffe.bin.dir/all] Error 2
Makefile:127: recipe for target 'all' failed
make[3]: *** [all] Error 2
CMakeFiles/openpose_caffe.dir/build.make:110: recipe for target 'caffe/src/openpose_caffe-stamp/openpose_caffe-build' failed
make[2]: *** [caffe/src/openpose_caffe-stamp/openpose_caffe-build] Error 2
CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/openpose_caffe.dir/all' failed
make[1]: *** [CMakeFiles/openpose_caffe.dir/all] Error 2
Makefile:83: recipe for target 'all' failed
make: *** [all] Error 2
```
原因是：ubuntu 16.04自带gcc版本是5.4的，与CUDA的一些配置冲突。另一方面，protubuf的版本是2.6的，不兼容。所以我们解决的办法是：
1. 修改cudn host_config.h配置文件`vim /usr/local/cuda/include/host_config.h`：
>\#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ > 9)
//#if __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ > 9)
 #error -- unsupported GNU version! gcc versions later than 4.9 are not supported!
 //#error -- unsupported GNU version! gcc versions later than 4.9 are not supported!
 #endif /* __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ > 9) */
 //#endif /* __GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ > 9) */
2. 卸载libprotoc 2.6.0 `sudo apt-get autoremove libprotobuf-dev protobuf-compiler`
3. 安装libprotoc 2.5.0[教程参考](https://blog.csdn.net/tdmyl/article/details/31811317)
4. ./configure --prefix=/usr/local/ CC=/usr/bin/gcc
最后编译通过！运行测试脚本。`./build/examples/openpose/openpose.bin --image_dir examples/media --display 0`返回正确结果！

gdp1368@outlook.com 's Blog

导航

最近发表

友情链接