jingweimo的个人博客分享 http://blog.sciencenet.cn/u/jingweimo

博文

Caffe for faster rcnn (Win10+VS2013+Cuda7.5+Matlab2016b)

已有 2873 次阅读 2018-6-11 02:50 |系统分类:科研笔记

  1. Envirionment setup

    Install vs2013 (it is said to be the only supported version, not confirmed), and then its compatiable cuda 7.5.

  2. Download Microsoft Caffe https://github.com/Microsoft/caffe (or https://github.com/BVLC/caffe/tree/windows), and unzip it

  3. Download faster rcnn for Matlab https://github.com/ShaoqingRen/faster_rcnn 

  4. Modidfy the CommonSettings.props file (located at the 'windows' subfolder) to accomodate and enable cuda, matlab and other tools involved.

2.png

 Note the CudaArchitecture setup (eg., 5.0 for GorForce 940M):

https://blog.csdn.net/JiaJunLee/article/details/52067962

https://blog.csdn.net/lishanlu136/article/details/51474582


 5.  Build the Caffe (located at the unzipped Mocrosoft Caffe folder, i.e., caffe-master) solution in Release mode:

     a.  set the propeprty of libcaffe:

3.png

    b. add the .cu, .cpp and .hpp files (located at the 'src' and 'include' files of the caffe-master folder) of roi_pooling_layer into the corresponding files of libcaffe.

    c. Build libcafe, caffe, matcaffe and others (building all the files would take a long time).


6.  Copy the generated Release folder to the 'external\caffe\matlab\caffe_faster_rcnn'

in the Matlab version faster rcnn folder.


 

7. Modify nvmex.m function (in the functions folder of Matlab faster rcnn):

6.png

7.png

8. Run faster_rcnn_build.m and startup.m (make sure Matlab is configured to use VS2013)

9. Run experiments/script_faster_rcnn_demo (https://github.com/ShaoqingRen/faster_rcnn#testing-demo)


Test results:

8.png


Sources:

http://www.cnblogs.com/maitre-zhao/p/6689953.html

https://blog.csdn.net/sinat_30071459/article/details/50546891

https://blog.csdn.net/zb1165048017/article/details/52767478

https://blog.csdn.net/hong__fang/article/details/51317424

https://blog.csdn.net/lalalazds/article/details/73863696

https://www.cnblogs.com/LaplaceAkuir/p/6445189.html

https://blog.csdn.net/mr_curry/article/details/54745116

https://blog.csdn.net/dzkd1768/article/details/69525610

https://blog.csdn.net/DoveJay/article/details/77948970 

https://blog.csdn.net/DoveJay/article/details/66972979

https://github.com/Microsoft/caffe

https://github.com/ShaoqingRen/faster_rcnn 



https://wap.sciencenet.cn/blog-578676-1118362.html

上一篇:Motion Control with LabView
下一篇:Updating Theano 0.9.0 to 1.02
收藏 IP: 35.22.98.*| 热度|

0

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...
扫一扫,分享此博文

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-5-14 11:56

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部