机器学习备注：Windows上运行Mask_RCNN模型

偏执的太偏执、 2022-05-10 01:48 184阅读 0赞

## 准备工作 ##

这些模型在Linux上运行比较方便，不过想起Windows便捷的各种工具，总是忍不住弄过来，

原模型的地址在这，  
https://github.com/matterport/Mask\_RCNN  
当然github上还有很多其他版本的，tensorflow, pytorch是主流，大家可自行搜索，安装使用过程大同小异。下面的是一个tensorflow的版本。

**注意：windows下的pycocotools版本和Linux不同**

需要用到的工具pycocotools在这  
不过要注意，这个pycocotools工具有Linux版和windows版本的，不要弄错了，官网链接给出的都是Linux版本的  
windows版本（ Windows下用这个 :-o ）  
https://github.com/philferriere/cocoapi  
linux版本，  
https://github.com/waleedka/coco

前提：已经安装好anaconda, 搭建好tensorflow开发环境  
提示：  
conda create -n tensorflow python=3.6  
activate tensorflow  
pip3 install (numpy, skimage, scipy, pillow, cython, etc......)

运行时中途碰到一个错误：  
ImportError: No module named cv2  
原因就是因为没有安装opencv-python，此时在tensorflow环境安装即可  
(tensorflow) (tensorflow) E:\\AtensorflowDev>pip3 install opencv-python

## 操作步骤如下 ##

step1 clone the Mask\_RCNN repository  
git clone [https://github.com/matterport/Mask\_RCNN][https_github.com_matterport_Mask_RCNN].git  
此时会在你的目录下生成一个Mask\_RCNN的文件夹，这个就是你项目的ROOT\_DIR，后面要用到  
我的目录是：E:\\AtensorflowDev\\Mask\_RCNN  
该目录下有一个requirements.txt文件，此时还需要运行一下  
pip3 install -r requirements.txt  
以确保所需要的包都安装好

step2 下载已经训练好的COCO weights (mask\_rcnn\_coco.h5)  
官网（[https://github.com/matterport/Mask\_RCNN/releases][https_github.com_matterport_Mask_RCNN_releases]）给出的地址是  
[https://github.com/matterport/Mask\_RCNN/releases/download/v2.1/mask\_rcnn\_balloon.h5][https_github.com_matterport_Mask_RCNN_releases_download_v2.1_mask_rcnn_balloon.h5]  
因为（）的原因，下载起来好慢，我下载了10多次，差不多花了一天，如果你运气和我一样差，那就用这个包吧  
链接：https://pan.baidu.com/s/12A0E6h-mP4zl9oqoTvj6Ig 密码：4eks

step3 把 mask\_rcnn\_coco.h5 放置到Mask\_RCNN这个目录下面  
E:\\AtensorflowDev\\Mask\_RCNN\\mask\_rcnn\_coco.h5

step4 安装coco工具  
$ pip install git+https://github.com/philferriere/cocoapi.git\#subdirectory=PythonAPI  
我的输出界面是这样的

1.（执行命令）
    (tensorflow) E:\AtensorflowDev>pip install git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI 
    (然后回车)
    2.（下面是运行后输出的结果，表示安装成功）
    Collecting git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI
      Cloning https://github.com/philferriere/cocoapi.git to c:\users\admini~1\appdata\local\temp\pip-req-build-a4gugnlj
    Building wheels for collected packages: pycocotools
      Running setup.py bdist_wheel for pycocotools ... done
      Stored in directory: C:\Users\ADMINI~1\AppData\Local\Temp\pip-ephem-wheel-cache-35xkjtfa\wheels\69\2b\12\2fa959e49f73d26cff202c2f4e5079096c9c57c8a8509fd75c
    Successfully built pycocotools
    Installing collected packages: pycocotools
    Successfully installed pycocotools-2.0

**如果你在安装pycocotools这一步碰到困难，请参考后面的pycocotools安装说明。**

step5 启动anaconda中的tensor开发环境，选择jupyter notebook，然后在jupyter中upload这个demo,  
E:\\AtensorflowDev\\MaskRCNN\\samples\\demo.ipynb

最后给出一张本人运行的结果全图，注意包括了一些调试信息，有点啰嗦，还有那个ROOT\_DIR路径，因为老是提示ModuleNotFoundError，我干脆写成绝对路径，另外说一下的是，我没有GPU，程序中的那个GPU\_COUNT=1仅仅是用来计算每个批次图片的张数（这里就一张）和真实的GPU数量无关  
![70][]

### pycocotools安装说明 ###

有的人可能会发现在线安装 比较费时间，有时我也是，一连试了好几次，安装不成功，不知道什么问题，止在这里，

(torch) E:\>pip install git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI
    Collecting git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI
      Cloning https://github.com/philferriere/cocoapi.git to c:\users\admini~1\appdata\local\temp\pip-req-build-mff3lz_x
    ...
    
    于是我的做法就是直接把包下载下来，本地编译安装，主要步骤就是切换到PythonAPI目录， 然后运行安装指令：
    python setup.py install
    编译运行之前请检查已经安装了VS2017工具（https://github.com/philferriere/cocoapi 上有说明）和cython。
    全部运行过程及内容如下，其中包括VS2017community工具使用的详细信息。(torch) E:\>cd E:\AMaskRCNN\pf-cocoapi-master\PythonAPI
    (torch) E:\AMaskRCNN\pf-cocoapi-master\PythonAPI>dir
     驱动器 E 中的卷是 文档
     卷的序列号是 84xx-xxxx
     E:\AMaskRCNN\pf-cocoapi-master\PythonAPI 的目录
    2018/10/24 周三  下午 04:00    <DIR>          .
    2018/10/24 周三  下午 04:00    <DIR>          ..
    2018/10/24 周三  下午 04:00    <DIR>          demos
    2018/10/24 周三  下午 04:00               199 Makefile
    2018/10/24 周三  下午 04:00    <DIR>          pycocotools
    2018/10/24 周三  下午 04:00               910 setup.py
                   2 个文件          1,109 字节
                   4 个目录 213,706,694,656 可用字节
    
    (torch) E:\AMaskRCNN\pf-cocoapi-master\PythonAPI>python setup.py install
    Compiling pycocotools/_mask.pyx because it changed.
    [1/1] Cythonizing pycocotools/_mask.pyx
    d:\Anaconda3\envs\torch\lib\site-packages\Cython\Compiler\Main.py:367: FutureWarning: Cython directive 'language_level' not set, using 2 for now (Py2). This will change in a later release! File: E:\AMaskRCNN\pf-cocoapi-master\PythonAPI\pycocotools\_mask.pyx
      tree = Parsing.p_module(s, pxd, full_module_name)
    running install
    running build
    running build_py
    creating build
    creating build\lib.win-amd64-3.7
    creating build\lib.win-amd64-3.7\pycocotools
    copying pycocotools\coco.py -> build\lib.win-amd64-3.7\pycocotools
    copying pycocotools\cocoeval.py -> build\lib.win-amd64-3.7\pycocotools
    copying pycocotools\mask.py -> build\lib.win-amd64-3.7\pycocotools
    copying pycocotools\__init__.py -> build\lib.win-amd64-3.7\pycocotools
    running build_ext
    building 'pycocotools._mask' extension
    creating build\temp.win-amd64-3.7
    creating build\temp.win-amd64-3.7\Release
    creating build\temp.win-amd64-3.7\Release\pycocotools
    creating build\temp.win-amd64-3.7\common
    C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\cl.exe /c /nologo /Ox /W3 /GL /DNDEBUG /MD -Id:\Anaconda3\envs\torch\lib\site-packages\numpy\core\include -I../common -Id:\Anaconda3\envs\torch\include -Id:\Anaconda3\envs\torch\include "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\cppwinrt" /Tcpycocotools/_mask.c /Fobuild\temp.win-amd64-3.7\Release\pycocotools/_mask.obj
    _mask.c
    d:\anaconda3\envs\torch\lib\site-packages\numpy\core\include\numpy\npy_1_7_deprecated_api.h(12) : Warning Msg: Using deprecated NumPy API, disable it by #defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
    pycocotools/_mask.c(3271): warning C4244: “=”: 从“Py_ssize_t”转换到“siz”，可能丢失数据
    pycocotools/_mask.c(3675): warning C4244: “函数”: 从“npy_intp”转换到“siz”，可能丢失数据
    pycocotools/_mask.c(6637): warning C4244: “=”: 从“npy_intp”转换到“siz”，可能丢失数据
    pycocotools/_mask.c(7383): warning C4244: “=”: 从“Py_ssize_t”转换到“siz”，可能丢失数据
    C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\cl.exe /c /nologo /Ox /W3 /GL /DNDEBUG /MD -Id:\Anaconda3\envs\torch\lib\site-packages\numpy\core\include -I../common -Id:\Anaconda3\envs\torch\include -Id:\Anaconda3\envs\torch\include "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\include" "-IC:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\include" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\ucrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\shared" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\um" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\winrt" "-IC:\Program Files (x86)\Windows Kits\10\include\10.0.17763.0\cppwinrt" /Tc../common/maskApi.c /Fobuild\temp.win-amd64-3.7\Release\../common/maskApi.obj
    maskApi.c
    C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\link.exe /nologo /INCREMENTAL:NO /LTCG /DLL /MANIFEST:EMBED,ID=2 /MANIFESTUAC:NO /LIBPATH:d:\Anaconda3\envs\torch\libs /LIBPATH:d:\Anaconda3\envs\torch\PCbuild\amd64 "/LIBPATH:C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\lib\x64" "/LIBPATH:C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\lib\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\lib\10.0.17763.0\ucrt\x64" "/LIBPATH:C:\Program Files (x86)\Windows Kits\10\lib\10.0.17763.0\um\x64" /EXPORT:PyInit__mask build\temp.win-amd64-3.7\Release\pycocotools/_mask.obj build\temp.win-amd64-3.7\Release\../common/maskApi.obj /OUT:build\lib.win-amd64-3.7\pycocotools\_mask.cp37-win_amd64.pyd /IMPLIB:build\temp.win-amd64-3.7\Release\pycocotools\_mask.cp37-win_amd64.lib
      正在创建库 build\temp.win-amd64-3.7\Release\pycocotools\_mask.cp37-win_amd64.lib 和对象 build\temp.win-amd64-3.7\Release\pycocotools\_mask.cp37-win_amd64.exp
    正在生成代码
    已完成代码的生成
    running install_lib
    creating d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    copying build\lib.win-amd64-3.7\pycocotools\coco.py -> d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    copying build\lib.win-amd64-3.7\pycocotools\cocoeval.py -> d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    copying build\lib.win-amd64-3.7\pycocotools\mask.py -> d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    copying build\lib.win-amd64-3.7\pycocotools\_mask.cp37-win_amd64.pyd -> d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    copying build\lib.win-amd64-3.7\pycocotools\__init__.py -> d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools
    byte-compiling d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools\coco.py to coco.cpython-37.pyc
    byte-compiling d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools\cocoeval.py to cocoeval.cpython-37.pyc
    byte-compiling d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools\mask.py to mask.cpython-37.pyc
    byte-compiling d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools\__init__.py to __init__.cpython-37.pyc
    running install_egg_info
    Writing d:\Anaconda3\envs\torch\Lib\site-packages\pycocotools-2.0-py3.7.egg-info
    
    (torch) E:\AMaskRCNN\pf-cocoapi-master\PythonAPI>

### 参考资料 ###

Object Detection and Segmentation in Python with Mask-RCNN  
http://bradsliz.com/2017-11-06-object-segmentation/  
[https://github.com/philferriere/cocoapi][https_github.com_philferriere_cocoapi]  
[https://github.com/matterport/Mask\_RCNN][https_github.com_matterport_Mask_RCNN]  
[https://github.com/waleedka/coco][https_github.com_waleedka_coco]

![70 1][]

AI视像算法学习群: 824991413

[https_github.com_matterport_Mask_RCNN]: https://github.com/matterport/Mask_RCNN
[https_github.com_matterport_Mask_RCNN_releases]: https://github.com/matterport/Mask_RCNN/releases
[https_github.com_matterport_Mask_RCNN_releases_download_v2.1_mask_rcnn_balloon.h5]: https://github.com/matterport/Mask_RCNN/releases/download/v2.1/mask_rcnn_balloon.h5
[70]: /images/20220510/4b160eb511a7454fb349585318ea775e.png
[https_github.com_philferriere_cocoapi]: https://github.com/philferriere/cocoapi
[https_github.com_waleedka_coco]: https://github.com/waleedka/coco
[70 1]: /images/20220510/f6942afe286a402ca08824cd884011b4.png