Convert vector to OpenCV Matd1d in c++ - c++

Question

I have a variable M that is a cv::Mat1d matrix. It is made like this:
cv::Mat1d<double> M;
It is populated with a bunch of values in some other code that is probably not necessary and it looks like this when I print it:
[-0.9344576352096885;
-0.9344576352096885;
-0.5766199600499906;
0.2846686026510846;
0.9589011777015718;
0.9285453673591227;
0.3137239980297359;
-0.2302718892981206;
-0.2921750731112262;
-0.2206633656711884;
-0.2175072323850435;
-0.1725991485554647;
-0.2140556050785325;
-0.4148403958730175;
-0.4036417215304363;
-0.06016889338878993;
0.3028103268622913;
0.4454375499811856;
0.3803583582813156;
0.3188387192279333;
0.3914868895364941;
0.4488724871465618;
0.2694705005556897;
-0.05248136866304744;
-0.2971598882254832;
-0.3545797186279719;
-0.2294426230118397;
-0.1673776410980104;
-0.2768386357175945;
-0.3276029287776189;
-0.2361695287135101;
-0.06139424097086685;
0.1621769468562924;
0.3275221571852822;
0.3153071221383847;
0.1341365194415481;
-0.04596232030582767;
-0.08961855126383761;
-0.02999823421387905;
-0.03225119207066054]
It is size [1 x 40].
I had to convert M to a vector X to do some calculations on it. I converted it like this:
vector<ld> X;
X.push_back(M.at<double>(0, 0));
for (int i = 1; i < M.rows; i++) {
X.push_back(M.at<double>(i, 0));
}
Now when I print it (After the calculations I did which are irrelevant), it looks like this:
[0 0 0 0 0 -1 -1 0 1 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 1 0 0]
and is size 40.
How do I convert X back to a [1x40] cv::Mat1d so that it is the same type as M?
Note: If my question/syntax/anything else is appalling, my apologies. I am just now leaving the relative safety of python for c++...
Edit
I think I solved it. In case anyone is curious OR if my solution needs correcting, here it is:
Mat1d test;
for (int i = 0; i < X.size(); i++) {
double new_val = (double) X[i];
test.push_back(new_val);
}

Answer 1

I'm not sure I quite understand what you need. You are doing two things here,
dealing with type conversion between integer and floating point types
reshaping the arrays between row and column vector
cv::Mat::reshape is a little different from numpy's reshape but does the same thing. you could have two Mats sharing the same data, and you use the one for a row view and the other for a column view.
cv::Mat m1_column, m2_row;
m2_row = m1_column.reshape(0, 1);
// first arg: number of channels. 0 means same.
// second arg: number of rows. make it a row vector.
cv::Mat::convertTo is roughly equivalent to numpy's ndarray.astype. it copies the values, converted to the desired type, into another cv::Mat.
cv::Mat m1_floats, m2_ints, m3_floats;
m1_floats.convertTo(m2_ints, CV_8U);
m2_ints.convertTo(m3_floats, CV_32F);
combine both as you need.
you don't even need to move between column and row format. just use a single index in Mat::at<>(). it'll run through the matrix as if it were flattened.
you can construct a Mat from a std::vector so it shares the vector's memory:
std::vector<int> X; // contains values from somewhere, ints should be 32 bit
cv::Mat X_as_a_Mat(1, X.size(), CV_32S, X.data()); // a single-row Mat sharing the std::vector's data
cv::Mat X_as_doubles;
X_as_a_Mat.convertTo(X_as_doubles, CV_64F); // converts data type
this code assumes that int is a 32 bit signed type (CV_32S). I don't know what element type your vector<ld> is supposed to be.