caffe openpose/Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields配置（转）
2017-08-07 10:33
621 查看
RealtimeMulti-Person2DPoseEstimationusingPartAffinityFields是CVPR2017的一篇论文，作者称是世界上第一个基于深度学习的实时多人二维姿态估计。优酷演示地址：链接前几天作者公布了windows下的代码，下面来说说如何配置：英文配置地址可以参考作者的github：https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/doc/installation.md#windows首先你得装好CUDA8，cudnn5.1，vs2015.其他版本都不行，只有vs2015可以装作者说要下载Python2.4.1364bits，其实我的anaconda2也可以（2.7的那个版本）cmake，ninja（我下载了，但是貌似没用到）下载windows版的openposehttps://github.com/CMU-Perceptual-Computing-Lab/openpose/tree/windows并解压，我放在D盘。下载model，作者给了2个，http://posefs1.perception.cs.cmu.edu/Users/tsimon/Projects/coco/data/models/coco/pose_iter_440000.caffemodel（精度高，速度较慢）下载后放到D:\openpose-windows\models\pose\coco下http://posefs1.perception.cs.cmu.edu/Users/tsimon/Projects/coco/data/models/mpi/pose_iter_160000.caffemodel（速度快，但是精度低）下载后放到D:\openpose-windows\models\pose\mpi\下打开D:\openpose-windows\3rdparty\caffe\caffe-windows\scripts下的build_win.cmd，开始编译。提示出错hash什么不匹配的，原因是网的问题，https://github-production-release-asset-2e65be.s3.amazonaws.com/39632178/8053cdd0-0068-11e7-8b60-c46f580c47e0?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAIWNJYAX4CSVEH53A%2F20170531%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20170531T072412Z&X-Amz-Expires=300&X-Amz-Signature=1860d1ab5ee6598f6a93e6230d1b74a09200e95e22d72021043f5176b9d73de3&X-Amz-SignedHeaders=host&actor_id=28419338&response-content-disposition=attachment%3B%20filename%3Dlibraries_v140_x64_py27_1.1.0.tar.bz2&response-content-type=application%2Foctet-stream下载后复制到D:\openpose-windows\3rdparty\caffe\dependencies\download下。再编译，等待半个小时，就会编译完毕。（若出现noccompilercouldbefound请参考我另一篇博文）接着打开D:\openpose-windows\windows_project\OpenPose.sln右键点击OpenPoseDemo生成，此时会报错。提示找不到D:\openpose-windows\3rdparty\lib\Releasecaffe.lib。其实它在D:\openpose-windows\3rdparty\caffe\lib\Release下，路径的问题。把release下的的两个文件放到D:\openpose-windows\3rdparty\lib\Release下就OK了。再编译又提示错误，说找不到pythoh27（其实也可以在build_win.cmd下把Python的路径改成你自己的python路径，这样这里就不会报错。）右键OpenPoseDemo，属性，连接器，常规，附加库目录，把你的python路径加进去，我为了防止出错，把2个路径都加了。再生成，就成功了。顺便再生成一下openpose（貌似没什么东西生成）此时你会在D:\openpose-windows\windows_project\x64\Release下找到OpenPoseDemo.exe。但是你双击它会提示你缺少各种东西。作者说把保存，并再次编译。发现oppenpose终于可以打开了！有些同学可能会提示报错outofmemory（比如说我），其实就是显存不够，爆显存了（1GGT650M伤不起啊）打开OpenPose.sln，双击OpenPoseDemo下的openpose.cpp.在第67行、68行修改网络的大小，我设定成如上的大小，1G显存也可以跑了。大功告成！如果想在视频下跑，可以在cmd下使用如下命令:windows_project\x64\Release\OpenPoseDemo.exe--videoexamples/media/video.avi如果是图片的话bin\OpenPoseDemo.exe--image_direxamples/media/
{openpose_folder}\3rdparty\caffe\caffe-windows\build\install\bin\下的所有DLL文件复制到{openpose_folder}\windows_project\x64\Release.下但是我没有找到这个路径。我把D:\openpose-windows\3rdparty\caffe\caffe-windows\scripts\build\tools\Release下的所有DLL文件复制过去。再把{}openpose_folder}\3rdparty\caffe\dependencies\libraries_v140_x64_py27_1.1.0\libraries\x64\vc14\bin\下的opencv_ffmpeg310_64.dll,
opencv_video310.dll和
opencv_videoio310.复制到{openpose_folder}\windows_project\x64\Release下。此时打开OpenPoseDemo.exe，发现还是缺少一些东西，我在D:\openpose-windows下搜索缺少的文件，并复制到{openpose_folder}\windows_project\x64\Release下。此时，发现oppenpose可以打开了，但是提示找不到model，其实是路径的问题，打开OpenPose.sln，双击OpenPoseDemo下的openpose.cpp.在第66行，手动设定models的绝对路径，比如我的是这个：
内容来自用户分享和网络整理，不保证内容的准确性，如有侵权内容，可联系管理员处理 
相关文章推荐
- [OpenPose翻译] Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields ∗原文翻译（注释版）
- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields ∗
- 论文阅读-Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields ∗ 实时多人人体姿态估计论文原理讲解
- 姿态论文整理--02-Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- 姿态估计论文思路整理 -- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- 行人姿态估计--Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- 论文阅读：《Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields》CVPR 2017
- 论文阅读笔记: 2017 cvpr Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- 拜读曹哲的Realtime Multi-Person 2D Pose Estimation using Part Affinity Fileds
- Code note: Realtime Multi-person 2D Pose estimation using Part Affinity Fields(2)
- READING NOTE: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- 论文阅读：RealTime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields 个人解读
- Paper Reading:Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Paper reading: Realtime Multi-person 2D Pose estimation using Part Affinity Fields(1)
- READING NOTE: Towards Accurate Multi-person Pose Estimation in the Wild
- Robust Real-Time Extreme Head Pose Estimation
- 论文阅读：《Towards accurate multi-person pose estimation in the wild》CVPR 2017