Build mnist dataset (dscompile) in Windows7 (VS2012)

[jong....@gmail.com]

Hi everyone,

I know developers not offer support for Windows, but I think someone successfully compile and use EBlearn in Windows environment.

I built EBLearn in Visual Studio 2012 (MSVC 11.0) at Windows 7.
After installing Boost_1_53_0, I compiled dscompile to build mnist data set in eblearn ( I know “mnist_compiled_data”)
I got this error:
_______________________________________________________________
Setting dataset name to: mnist_train
Warning: pid() not implemented for Windows
Error: Boost libraries not available, install libboost-filesystem-dev libboost-regex-dev and recompile, in ebl::dataset<float>::dataset at c:\project\eblearn\tools\libeblearntools\include\dataset.
147
_______________________________________________________________

I checked that Boost libs: filesystem, regex, system, python, thread were successfully built.
Therefore, I comment BOOST_LIB_ERROR in dataset.hpp as followings :
#ifndef __BOOST__
// eblerror(BOOST_LIB_ERROR);
#endif

Then, I got these messages:
_______________________________________________________________

Dataset compiler for libeblearn library
_______________________________________________________________
input parameters:
dataset name: mnist_train
dataset type: regular
dataset precision: float
images root directory: mnist rain
annotations directory:
ignored annotations directory:
output directory: eb_dataset
outputs: eb_dataset/mnist_train_*.mat
images pattern: .*[.](png|jpg|jpeg|PNG|JPG|JPEG|bmp|BMP|ppm|PPM|pnm|PNM|pgm|PGM|gif|GIF|mat|MAT)
channels mode: RGB
preprocessing: yes
display: no
display sleep: 0 ms.
shuffling: no
usepose: no
useparts: no
partsonly: no
stereo: no
max per class limitation: none
max data limitation: none
mexican_hat_size: 0
preprocessing kernel size: [ 7x7 ]
deformations: -1
resizing method: bilinear
output dimensions: [ 32x32x1 ]
minimum input dimensions: <empty>
maximum input dimensions: <empty>
no padded: no
scales: none
laplacian pyramid: 0
_______________________________________________________________
Setting dataset name to: mnist_train
Warning: pid() not implemented for Windows
.
Setting preprocessing to: 7x7_bilinear32x32
Preprocessing produces 1 layers per sample.
Preprocessing modules:
0: resizepp module 7x7_bilinear32x32, resizing with method 0 to 32x32 while preserving aspect ratio, pp: rgb_to_rgb
Setting output directory to eb_dataset
Setting temporary output directory to eb_dataset
Setting target dimensions to 32x32x1
Setting saving mode to: dynset
Enabling individual sample saving.
Disabling saving sample layers separately.
Setting image pattern to .*[.](png|jpg|jpeg|PNG|JPG|JPEG|bmp|BMP|ppm|PPM|pnm|PNM|pgm|PGM|gif|GIF|mat|MAT)
Found: 0 total samples.
Error: no samples found, in ebl::dataset<float>::count_total at c:\project\eblearn\tools\libeblearntools\include\dataset.hpp:1507
Press any key to continue . . .
------------------------------------------------------------------

Anyone helps me to run dscompile in Windows 7 (MSVC 11.0)
Thanks in advance

[jong....@gmail.com]

I found and fixed error. It was from boost path problem.

[nguyenti...@gmail.com]

Hi,

Could you please let me know which external dependencies are needed to build eblearn for MS visual studio?