Build caffe, pycaffe and matcaffe on Windows

Requirements:

Download and install the following software's:
  1. Git for Windows: https://git-scm.com/download/win
  2. Visual Studio 2015: https://www.visualstudio.com/downloads/. To download older versions of visual studio you can go to https://www.visualstudio.com/vs/older-downloads/ 
  3. Cmake 3.4 or higher https://cmake.org/download/
  4. Python 3.5 Anaconda x64 (Required to build pycaffe). https://repo.continuum.io/archive/Anaconda3-4.2.0-Windows-x86_64.exe
  5. CUDA 8.0 https://developer.nvidia.com/accelerated-computing-toolkit
  6. cuDNN v5.1 (optional) https://developer.nvidia.com/cudnn
  7. Matlab (Required to build matcaffe)  http://mathworks.com

Instructions:

  1. Install Cmake and add it to system path.
  2. Install CUDA v8.0 (For GPU support only)
  3. Extract cuDNN (For GPU Support only)
    • copy all files from bin folder to 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin'
    • similarly from include to 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\include' 
    • and 'lib/x64' to  ''C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib/x64'.
  4. Optional - Install Anaconda, make sure you select Register Anaconda as system path for Python 3.5 (required to build pycaffe).
  5. Optional - Install and activate Matlab using your Matlab licence (required to build matcaffe).
  6. Open command window in the desired folder "D:\Downloads\caffe" (Shift+RightClick and Open Command Window here).
  7. Enter the following commands
    D:\Downloads\caffe> git clone https://github.com/BVLC/caffe.git
    D:\Downloads\caffe> cd caffe
    D:\Downloads\caffe\caffe> git checkout windows
  8. Open scripts folder and edit build_win.cmd
    • set WITH_NINJA=0
    • set PYTHON_VERSION=3
    • set CONDA_ROOT to Anaconda folder, "C:\Program Files\Anaconda3".
    • set BUILD_MATLAB=1
    • Add -CUDNN_ROOT="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0" in cmake configuration in line 170.
  9. Save build_win.cmd and enter following command in command window
    D:\Downloads\caffe\caffe> scripts\build_win.cmd
  10. After successful build enter the following commands to run caffe
    D:\Downloads\caffe\caffe> cd build\tools\Release
    D:\Downloads\caffe\caffe\build\tools\Release> caffe -help
  11. Pycaffe instructions:

  12. Enter the following commands
    D:\Downloads\caffe\caffe\build\tools\Release> cd ..\..\..\
    D:\Downloads\caffe\caffe> conda config --add channels conda-forge
    D:\Downloads\caffe\caffe> conda install ninja --yes
  13. Copy the contents of D:\Downloads\caffe\caffe\python to site-packages folder [C:\Program Files\Anaconda3\]Lib\site-packages.
  14. Now run the following commands:
    D:\Downloads\caffe\caffe> python
    >>> import caffe
  15. You should be able to successfully import and run pycaffe.
  16. Matcaffe instructions:

  17. Enter the following command to download caffe bvlc model
    D:\Downloads\caffe\caffe> python scripts\download_model_binary.py models\bvlc_reference_caffenet
  18. Copy caffe_.mexw64 from D:\Downloads\caffe\caffe\matlab\+caffe\private\Release\ to D:\Downloads\caffe\caffe\matlab\+caffe\private.
  19. Add D:\Downloads\caffe\caffe\build\libraries to Matlab Path using setenv() command.
  20. Now you should be able to run classification_demo.m in C:\Projects\caffe\matlab\demo folder.

No comments:

Post a Comment

How to run R code in Python

R is freely available language and computational tool which is very popular among the statisticians and bioinformatics community. In this p...