How to convert Opencv Mat to Alglib real 2D array?
Here is an example where I am stucked
Mat Col(28539,97,CV_32F);
I want to convert this Mat to alglib real_2d_array for training a classifier.
Mat Col(28539, 97, CV_32F);
is a OpenCV bi-dimensional (28539 rows, 97 columns) dense floating-point (CV_32F = float) array.
The alglib almost-equivalent datatype is
// bi-dimensional real (double precision) array
real_2d_array matrix;
The data layout in Mat is compatible with real_2d_array (and the majority of dense array types from other toolkits and SDKs).
A simple way to convert is:
const int rows(28539);
const int columns(97);
matrix.setlength(rows, columns);
for (int i(0); i < rows; ++i)
for (int j(0); j < columns; ++j)
matrix(i, j) = Col.at<float>(i, j);
Mat::at returns a reference to the specified array element.
EDIT
From the reference manual:
void alglib::dfbuildrandomdecisionforest(
real_2d_array xy,
ae_int_t npoints,
ae_int_t nvars,
ae_int_t nclasses,
ae_int_t ntrees,
double r,
ae_int_t& info,
decisionforest& df,
dfreport& rep);
xy is the training set (lines corresponding to sample components and columns corresponding to variables).
For a classification task the first nvars of the columns contain independent variables. The last column will contain the class number (from 0 to nclasses-1). Fractional values are rounded to the nearest integer.
npoints is the training set size (>=1).nvars is the number of independent variables (>=1).nclasses must be >1 for classification.ntrees is the number of trees in a forest (>=1).r is the percent of a training set used to build individual trees (0 < R <= 1).The remaining parameters are output parameters. In case of problems you should check info:
info return code:
[0..nclasses-1].npoints<1, nvars<1, nclasses<1, ntrees<1, r<=0 or r>1).