adikhosla/feature-extraction

Computer vision feature extraction toolbox for image classification

feature-extraction's Issues

Info about Classifiers

Hi,
I have seen that in the demo.m file there is a part about the svm if i uncomment it doesn't work, giving me :Length of label vector does not match # of instances.

compile errors

Hi
When I try to extract lbp features and ssim features with Matlab2015b, the code cannot work.
For lbp feature，the errors are listed below:

Dataset: demo
Processing filelist (train, lbp): batch 0 of 1
Processing filelist (train, lbp): batch 1 of 1
Error using getmapping (line 37)
ASSUMEDTYPE must be an integer type name.

Error in lbp_original (line 3)
mapping=getmapping(8,'u2');

Error in lbp_original_4x4 (line 3)
feature_L0(:,1) = lbp_original(I);

Error in lbp_feature (line 10)
[lbp_L0 lbp_L1 lbp_L2] = lbp_original_4x4(im);

Error in extract_lbp (line 17)
d = lbp_feature(img);

Error in extract_feature (line 15)
[feat, x, y, wid, hgt] = feval(['extract_' feature], img, c);

Error in filelist_feature (line 23)
parfor i=1:length(filelist)

Error in batch_feature (line 42)
poolfeat = filelist_feature('', filelist(this_batch), feature, c);

Error in datasets_feature (line 34)
batch_feature(train_lists{i}, 'train', feature, c);

Error in demo (line 24)
datasets_feature(datasets, train_lists, test_lists, feature, c);

For ssim featurem, the errors are listed below:

Dataset: demo
Learning dictionary for feature: ssim, size 20
Running k-means, dictionary size 20...done!
Saving dictionary: cache//demo//dictionary_ssim_20.mat
Processing filelist (train, ssim): batch 0 of 1
Processing filelist (train, ssim): batch 1 of 1
Error using cellfun
Non-scalar in Uniform output, at index 1, output 1.
Set 'UniformOutput' to false.

Error in max_pooling (line 36)
emptyIdx = find(cellfun(@(x) sum(x)==0, idx));

Error in filelist_feature (line 37)
poolfeat = max_pooling(llcfeat, info, c.pool_region, p.pyramid_levels);

Error in batch_feature (line 42)
poolfeat = filelist_feature('', filelist(this_batch), feature, c);

Error in datasets_feature (line 34)
batch_feature(train_lists{i}, 'train', feature, c);

Error in demo (line 24)
datasets_feature(datasets, train_lists, test_lists, feature, c);

Demo not working in Octave

I am interested to do some personal work in Image processing and for this I need Image feature extraction. So this project interests me as a starting point. I am new to Octave, but I have used MatLab in past. I am trying to get the demo to work, for this I did the following (version details are also provided):

$ git clone [email protected]:adikhosla/feature-extraction.git
$ cd feature-extraction/
$ octave
octave:1> demo
warning: load: file found in load path
warning: function name 'config_hog3x3' does not agree with function file name '/home/imyousuf/projects/feature-extraction/features/ssim/config_ssim.m'
Dataset: demo
Learning dictionary for feature: color, size 20
error: default_formats: /usr/lib/x86_64-linux-gnu/octave/3.8.2/oct/x86_64-pc-linux-gnu/__magick_read__.oct: failed to load: /usr/lib/x86_64-linux-gnu/octave/3.8.2/oct/x86_64-pc-linux-gnu/__magick_read__.oct: undefined symbol: _ZN6Magick5ColorC1Ehhh
error: called from:
error:   /usr/share/octave/3.8.2/m/image/imformats.m at line 255, column 11
error:   /usr/share/octave/3.8.2/m/image/imformats.m at line 78, column 3
error:   /usr/share/octave/3.8.2/m/image/private/imageIO.m at line 60, column 7
error:   /usr/share/octave/3.8.2/m/image/imread.m at line 107, column 30
error:   /home/imyousuf/projects/feature-extraction/util/imgread.m at line 12, column 7
error:   /home/imyousuf/projects/feature-extraction/util/build_dictionary.m at line 49, column 11
error:   /home/imyousuf/projects/feature-extraction/util/datasets_feature.m at line 31, column 45
error:   /home/imyousuf/projects/feature-extraction/demo.m at line 24, column 1
octave:1> ver
----------------------------------------------------------------------
GNU Octave Version 3.8.2
GNU Octave License: GNU General Public License
Operating System: Linux 3.19.0-30-generic #33-Ubuntu SMP Mon Sep 21 20:58:04 UTC 2015 x86_64
----------------------------------------------------------------------
Package Name  | Version | Installation directory
--------------+---------+-----------------------
     control *|   2.6.6 | /usr/share/octave/packages/control-2.6.6
     general *|   1.3.4 | /usr/share/octave/packages/general-1.3.4
       image *|   2.2.2 | /usr/share/octave/packages/image-2.2.2
          io *|   2.2.4 | /usr/share/octave/packages/io-2.2.4
      signal *|   1.3.0 | /usr/share/octave/packages/signal-1.3.0
  statistics *|   1.2.3 | /usr/share/octave/packages/statistics-1.2.3

Please help me to get it to work in Octave. Thank you.

Octave compatibility

I tried to run it on octave without much success yet.

The following issues I noticed, most of them I already fixed while keeping the Matlab function as before. Pull request will be made soon as soon as I get demo.m working

1. Can't compile and getting this error from conf.m:
   error: 'RandStream' undefined near line 51 column 10
   error: called from:
   error: feature-extraction/util/conf.m at line 51, column 8
   error: feature-extraction/demo.m at line 20, column 3
2. Octave doesn't have the RandStream function
3. There is no matlabpool in Octave
4. feature-extraction/util/batch_feature.m at line 59, column error: save: Unrecognized option '-v7.3'

Compiling Code

Hi, I am trying to run your code in Matlab R2011a and having quite a bit of difficulty. Compile.m gives me several error messages such as "features/ssim/vggSsim/MEXfindnearestl2.cpp: In function 'void mexFunction(int, mxArray**, int, const mxArray**)':
features/ssim/vggSsim/MEXfindnearestl2.cpp:62: warning: deprecated conversion from string constant to 'char*'"

Then when I try demo.m, I get this type of error:
??? Undefined function or method 'matlabpool' for input arguments of type 'char'.

Error in ==> openPool at 20
if(matlabpool('size')==0)

Error in ==> datasets_feature at 16
openPool(c.cores);

Error in ==> demo at 24
datasets_feature(datasets, train_lists, test_lists, feature, c);

Could you please help?

Thanks,

Problem when compiling the program

Respected Sir,
When I run following command
datasets_feature(datasets, train_lists, test_lists, feature, c);
I got error:
Error using imread (line 349)
File "sun2.jpg" does not exist.

Error in imgread (line 12)
I = imread(img);

Error in build_dictionary>(parfor body)
(line 49)
img = imgread(filelist{perm(i)},
p);

Error in build_dictionary (line 47)
parfor i=1:num_images

Error in datasets_feature (line 31)
c.feature_config.(feature).dictionary
= build_dictionary(train_lists{i},
feature, c);

how to remeove this error.
Thanks

Error compile( Error using mex)

Good day!
When starts compile.m , I have compile error:

>> compile
Building with 'MinGW64 Compiler (C++)'.
Error using mex
D:\antispoofing\Matlab_spoofdet_hog_lbp_Gab__svm\feature-extraction-master\util\pixelwise_hog31.cc: In function 'mxArray* process(const mxArray*, const mxArray*)':
D:\antispoofing\Matlab_spoofdet_hog_lbp_Gab__svm\feature-extraction-master\util\pixelwise_hog31.cc:48:44: error: cannot convert 'const size_t* {aka const long long unsigned int*}' to 'const int*' in initialization
const int dims = mxGetDimensions(mximage);
^
D:\antispoofing\Matlab_spoofdet_hog_lbp_Gab__svm\feature-extraction-master\util\pixelwise_hog31.cc:81:72: error: cannot convert 'int' to 'const size_t* {aka const long long unsigned int*}' for argument '2' to 'mxArray*
mxCreateNumericArray_730(size_t, const size_t*, mxClassID, mxComplexity)'
mxArray *mxfeat = mxCreateNumericArray(3, out, mxDOUBLE_CLASS, mxREAL);

getmapping.m needs to be updated

The getmapping.m function needed for LBP extraction has been updated to version 0.2 with fixes to work in Matlab 2012 or newer

Meaning of 'dictionary_size'?

Could anyone please explain the meaning of 'dictionary_size' and how it affects the speed? Thank you.

Would you please implement and include the histogram refinement explained in https://github.com/Ashwani21/Local-texture-descriptors in this repo.

Histogram refinement results in https://github.com/Ashwani21/Local-texture-descriptors are very promising.

Warning: matlabpool will be removed in a future release.

Hi, adikhosla
I run the feature-extraction in matlab and the version is MATALB 2014a. When I run the demo script, it throws the warnning:

Warning: matlabpool will be removed in a future release.
To query the size of an already started parallel pool, query the 'NumWorkers' property of the pool.
To check if a pool is already started use 'isempty(gcp('nocreate'))'. 

The warning come from the file:

function [] = openPool(s)
%
% Copyright Aditya Khosla http://mit.edu/khosla
%
% Please cite this paper if you use this code in your publication:
%   A. Khosla, J. Xiao, A. Torralba, A. Oliva
%   Memorability of Image Regions
%   Advances in Neural Information Processing Systems (NIPS) 2012
%

if(matlabpool('size')==0)
    if(s==0)
        matlabpool;
    elseif(s>1)
        matlabpool(s);
    end
end

It seems the matlabpool leads the warning. Though it's not a problem, but the warning makes one feel bad.