I'm currently cross-compiling on the BeagleBone Black in a Visual Studio environment using Armadillo to translate MATLAB code into C++.
This is a signal processing project, so I need a way to read and write binary data files, specifically .mat files. Thankfully, the armadillo documentation says that you can load .mat files directly into a matrix using .load()
I attempted that at first, but it seems like it's not reading the file correctly, nor is it reading all the entries. My reference file is a 2000x6 matrix, and the created armadillo matrix is 5298x1. I know that without an armadillo-mimicking header, it will be converted into a column vector and I will need to reshape it using .reshape(), yet it simply isn't receiving all the entries, and by inspection, the entries it did read are wrong.
I'm not sure what the problem is. I've placed the data reference .mat files in the Debug folder for the remote project on the BBB, where the .out compiled file is created. Is there another way I should integrate it?
Also, help with mimicking the armadillo header or other suggestions are welcome. If you need anything, please let me know.
Here is the test program I am using:
#include <iostream>
#include <armadillo>
using namespace std;
using namespace arma;
int main()
{
mat data_ref;
data_ref.load("Epoxy_6A_Healthy_Output_200kHz_Act1_001.mat");
cout << "For Data_ref, there are " << data_ref.n_cols << " columns and " << data_ref.n_rows << " rows.\n";
cout << "First item: " << data_ref(0) << "\n6th item: " << data_ref(6) << "\n2000th item: " << data_ref(2000);
data_ref.reshape(2000, 6);
cout << "For Data_ref, there are " << data_ref.n_cols << " columns and " << data_ref.n_rows << " rows.\n";
cout << "First item: " << data_ref(0,0) << "\nLast Item: " << data_ref(1999,5);
cout << "\nDone";
return 0;
}
The first element in the .mat file is 0.0, and the last element is 0.0014. Here is the output.
For Data_ref, there are 1 columns and 5298 rows.
First item: 8.48749e-53
th item: 9.80727e+256
th item: -2.4474e+238For Data_ref, there are 6 columns and 2000 rows.
First item: 8.48749e-53
(gdb) 1028-var-list-children --simple-values "var4.public" 0 1000
(gdb) 1030-var-list-children --simple-values "var4.arma::Base<double,
arma::Mat<double> >" 0 1000
Last Item: 0
Done=thread-exited,id="1",group-id="i1"
The program '' has exited with code 0 (0x0).
Thanks
Armadillo does not support Matlab's .mat format. In the documentation they refer to the Armadillo
mat
binary format. You may however save the data in Matlab using thehdf5
binary format and import it into Armadillo but then you have to download the hdf5 lib and reconfigure Armadillo. See thehdf5_binary
section in the documentation.