opencv2 covariance matrix strange results - opencv

The following code gives inconsistent covariance matrix sizes.
cv::Mat A = (cv::Mat_<float>(3,2) << -1, 1, -2, 3, 4, 0);
cv::Mat covar1, covar2, covar3, covar4, mean;
calcCovarMatrix(A, covar1, mean, CV_COVAR_NORMAL | CV_COVAR_ROWS);
calcCovarMatrix(A, covar2, mean, CV_COVAR_SCRAMBLED | CV_COVAR_ROWS);
calcCovarMatrix(A, covar3, mean, CV_COVAR_NORMAL | CV_COVAR_COLS);
calcCovarMatrix(A, covar4, mean, CV_COVAR_SCRAMBLED | CV_COVAR_COLS);
std::cout << "size: " << covar1.size() << "\n";
std::cout << "size: " << covar2.size() << "\n";
std::cout << "size: " << covar3.size() << "\n";
std::cout << "size: " << covar4.size() << "\n";
covar1 and covar2 should have the same size because they both describe the covariance over the rows, and covar3 and covar4 should have the same size because they both describe the covariance over the columns, respectively. However, the output is:
size: [2 x 2]
size: [3 x 3]
size: [3 x 3]
size: [2 x 2]

The calcCovarMatrix() docs, specifically say that when using CV_COVAR_SCRAMBLED "The covariance matrix will be nsamples x nsamples."

Related

getline, stringstream, delimiter stuff

how do I use getline to separate a string ex. "you8only8live8once"? This solution only sets my first word to "you" and then stops at the first "8".
Also if I want to the "8" to be an arbitrary delimiter do I just put in cin >> delimiter; then change "getline(cin, phrase, '8');" to "getline(cin, phrase, delimiter);"?
istringstream inSS;
string phrase;
string firstWord;
string secondWord;
string thirdWord;
string fourthWord;
cout << "Please enter a digit infused string to explode:" << endl;
cout << "Please enter the digit delimiter:" << endl;
getline(cin, phrase, '8');
inSS.str(phrase);
inSS >> firstWord >> secondWord >> thirdWord >> fourthWord;
cout << "The 1st word is: " << firstWord << endl;
cout << "The 2nd word is: " << secondWord << endl;
cout << "The 3rd word is: " << thirdWord << endl;
cout << "The 4th word is: " << fourthWord << endl;
return 0;

Can't Figure out Logic Error involving Setting Rotation Matrix

I'm trying to extract the 3x3 rotation matrix from the 3x4 pose matrix I have. However, two values are differing even though I have very simple code setting one to the other. I'm banging my head against the wall because I have no idea why this is happening. Here is the code:
std::cout << "Camera pose matrix from optical flow homography" << std::endl;
for (int e = 0; e < pose.rows; e++) {
for (int f = 0; f < pose.cols; f++) {
std::cout << pose.at<double>(e,f) << " " << e << " " << f;
std::cout << " ";
}
std::cout << "\n" << std::endl;
}
std::cout << "Creating rotation matrix" << std::endl;
Mat rotvec = Mat::eye(3, 3, CV_32FC1);
for (int s = 0; s < pose.rows; s++) {
for (int g = 0; g < pose.cols-1; g++) {
rotvec.at<double>(s, g) = pose.at<double>(s,g);
std::cout << rotvec.at<double>(s,g) << " " << s << " " << g;
std::cout << " ";
}
std::cout << "\n" << std::endl;
}
std::cout << "Rotation matrix" << std::endl;
for (int e = 0; e < pose.rows; e++) {
for (int f = 0; f < pose.cols-1; f++) {
std::cout << rotvec.at<double>(e,f) << " " << e << " " << f;
std::cout << " ";
std::cout << pose.at<double>(e,f) << " " << e << " " << f;
std::cout << " ";
}
std::cout << "\n" << std::endl;
}
Here is the output:
Camera pose matrix from optical flow homography
5.26354e-315 0 0 0 0 1 0.0078125 0 2 0 0 3
0.0078125 1 0 0 1 1 0 1 2 5.26354e-315 1 3
0 2 0 5.26354e-315 2 1 1.97626e-323 2 2 7.64868e-309 2 3
Creating rotation matrix
5.26354e-315 0 0 0 0 1 0.0078125 0 2
0.0078125 1 0 0 1 1 0 1 2
0 2 0 5.26354e-315 2 1 1.97626e-323 2 2
Rotation matrix
5.26354e-315 0 0 5.26354e-315 0 0 0 0 1 0 0 1 5.26354e-315 0 2 0.0078125 0 2
0.0078125 1 0 0.0078125 1 0 0 1 1 0 1 1 0.0078125 1 2 0 1 2
0 2 0 0 2 0 5.26354e-315 2 1 5.26354e-315 2 1 1.97626e-323 2 2 1.97626e-323 2 2
Here you can see I'm trying to save the first three columns of pose into the rotvec matrix. When I actually set the rotation matrix equal to the pose for those three columns, I get the correct matrix, as the second matrix is equal to the first three columns of the first matrix. However, when I check the rotation matrix once again, (third matrix) it is not the same as the output I require on coordinates (0, 2) and (1, 2). (I outputted the rotvec matrix number next to the pose matrix number, and you can see at these coordinates the numbers do not match). I am not sure why this is happening, could someone please help me out?
Solved my problem for anyone else who stumbles upon this later: I just changed the Mat type to CV_64F (to make it double) for both rotvec and pose, and used all to display. Creds to berak for pointing me in the right direction.

implications of using _mm_shuffle_ps on integer vector

SSE intrinsics includes _mm_shuffle_ps xmm1 xmm2 immx which allows one to pick 2 elements from xmm1 concatenated with 2 elements from xmm2. However this is for floats, (implied by the _ps , packed single). However if you cast your packed integers __m128i, then you can use _mm_shuffle_ps as well:
#include <iostream>
#include <immintrin.h>
#include <sstream>
using namespace std;
template <typename T>
std::string __m128i_toString(const __m128i var) {
std::stringstream sstr;
const T* values = (const T*) &var;
if (sizeof(T) == 1) {
for (unsigned int i = 0; i < sizeof(__m128i); i++) {
sstr << (int) values[i] << " ";
}
} else {
for (unsigned int i = 0; i < sizeof(__m128i) / sizeof(T); i++) {
sstr << values[i] << " ";
}
}
return sstr.str();
}
int main(){
cout << "Starting SSE test" << endl;
cout << "integer shuffle" << endl;
int A[] = {1, -2147483648, 3, 5};
int B[] = {4, 6, 7, 8};
__m128i pC;
__m128i* pA = (__m128i*) A;
__m128i* pB = (__m128i*) B;
*pA = (__m128i)_mm_shuffle_ps((__m128)*pA, (__m128)*pB, _MM_SHUFFLE(3, 2, 1 ,0));
pC = _mm_add_epi32(*pA,*pB);
cout << "A[0] = " << A[0] << endl;
cout << "A[1] = " << A[1] << endl;
cout << "A[2] = " << A[2] << endl;
cout << "A[3] = " << A[3] << endl;
cout << "B[0] = " << B[0] << endl;
cout << "B[1] = " << B[1] << endl;
cout << "B[2] = " << B[2] << endl;
cout << "B[3] = " << B[3] << endl;
cout << "pA = " << __m128i_toString<int>(*pA) << endl;
cout << "pC = " << __m128i_toString<int>(pC) << endl;
}
Snippet of relevant corresponding assembly (mac osx, macports gcc 4.8, -march=native on an ivybridge CPU):
vshufps $228, 16(%rsp), %xmm1, %xmm0
vpaddd 16(%rsp), %xmm0, %xmm2
vmovdqa %xmm0, 32(%rsp)
vmovaps %xmm0, (%rsp)
vmovdqa %xmm2, 16(%rsp)
call __ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc
....
Thus it seemingly works fine on integers, which I expected as the registers are agnostic to types, however there must be a reason why the docs say that this instruction is only for floats. Does someone know any downsides, or implications I have missed?
There is no equivalent to _mm_shuffle_ps for integers. To achieve the same effect in this case you can do
SSE2
*pA = _mm_shuffle_epi32(_mm_unpacklo_epi32(*pA, _mm_shuffle_epi32(*pB, 0xe)),0xd8);
SSE4.1
*pA = _mm_blend_epi16(*pA, *pB, 0xf0);
or change to the floating point domain like this
*pA = _mm_castps_si128(
_mm_shuffle_ps(_mm_castsi128_ps(*pA),
_mm_castsi128_ps(*pB), _MM_SHUFFLE(3, 2, 1 ,0)));
But changing domains may incur bypass latency delays on some CPUs. Keep in mind that according to Agner
The bypass delay is important in long dependency chains where latency is a bottleneck, but
not where it is throughput rather than latency that matters.
You have to test your code and see which method above is more efficient.
Fortunately, on most Intel/AMD CPUs, there is usually no penalty for using shufps between most integer-vector instructions. Agner says:
For example, I found no delay when mixing PADDD and SHUFPS [on Sandybridge].
Nehalem does have 2 bypass-delay latency to/from SHUFPS, but even then a single SHUFPS is often still faster than multiple other instructions. Extra instructions have latency, too, as well as costing throughput.
The reverse (integer shuffles between FP math instructions) is not as safe:
In Agner Fog's microarchitecture on page 112 in Example 8.3a, he shows that using PSHUFD (_mm_shuffle_epi32) instead of SHUFPS (_mm_shuffle_ps) when in the floating point domain causes a bypass delay of four clock cycles. In Example 8.3b he uses SHUFPS to remove the delay (which works in his example).
On Nehalem there are actually five domains. Nahalem seems to be the most effected (the bypass delays did not exist before Nahalem). On Sandy Bridge the delays are less significant. This is even more true on Haswell. In fact on Haswell Agner said he found no delays between SHUFPS or PSHUFD (see page 140).

Issue with cv::Mat::zeros initialization

my problem is just astonishing. This is the code
#define NCHANNEL 3
#define NFRAME 100
Mat RR = Mat::zeros(NCHANNEL, NFRAME-1, CV_64FC1);
double *p_0 = RR.ptr<double>(0);
double *p_1 = RR.ptr<double>(1);
double *p_2 = RR.ptr<double>(2);
cout<< p_0[NFRAME-1] << endl << p_1[NFRAME-1] << endl << p_2[NFRAME-1] << endl;
And the output is: 0 0 -6.27744e+066 .
Where is that awful number come from? it seems I'm printing a pointer or something rough in memory. (uh, 0 is the value of all other elements, of course).
You are accessing after the last element of Mat. If you use NFRAME-1 for initialization then the last element has NFRAME-2 index.

setw() - adjustfield (left, right or internal)

I am going though the string functions doing tests to learn them (I am a newbie programmer)
Anyway, I am currently looking at setw() but I seam to not understand it... I think I understand the basic use and the use of setfil
here is my test code
http://ideone.com/czAXH
Anyway the cplusplus website says.. "format flag adjustfield (left, right or internal)" but doesn't say how to use this?
I assume this means I can do the above code but place the "spacing" after the word instead of before it..
How do I do that?
std::cout << std::left
<< "[" << std::setw(3) << 1 << "," << std::setw(5) << -100 << "]\n";
std::cout << std::internal
<< "[" << std::setw(3) << 1 << "," << std::setw(5) << -100 << "]\n";
std::cout << std::right
<< "[" << std::setw(3) << 1 << "," << std::setw(5) << -100 << "]\n";
Outputs:
[1 ,-100 ]
[ 1,- 100]
[ 1, -100]

Resources