larry37's profile - activity

Apart from core and imgproc, building also requires hal, zlib and ippicv libraries. The first two are small, but the Intel IPP is 35MB. Could this be the reason it's so large? Any options other than compiling without IPP?

Building with statically linked OpenCV 2.49 previously added ~1MB to my executable size. Building the same application with OpenCV 3.0 adds 10MB to the final executable. Is there a reason for the 10x as large output? The Core and ImgProc libraries that I'm using don't seem to have changed much in size.

Thanks for the quick answer, worked perfectly.

I'm having trouble migrating BRIEF feature extraction to OpenCV 3. I have followed the transition guide from http://docs.opencv.org/master/db/dfa/... to compile the contrib module and load it into opencv-master. I am including the <xfeatures2d.hpp> file and trying to get descriptors with the following code (which worked fine with OpenCV2):

cv::xfeatures2d::BriefDescriptorExtractor extractor;
extractor.create(64);

cv::KeyPoint kp = cv::KeyPoint(x, y, size);
std::vector<cv::KeyPoint> kp_vec();
kp_vec.push_back(kp);
cv::Mat descriptors;

extractor.compute(image, kp_vec, descriptors);

The program crashes on the last line inside features2d.h, with the error CV_Error(Error::StsNotImplemented, "");. Am I doing something wrong? Is the BRIEF detector missing? I could not find any examples in the openCV docs, apart from some python code that looks like the one I posted above.

Opencv's findFundamentalMat gives different results from the one returned by stereoCalibrate, on the same points (10 checkerboards worth of points concatenated in a long vector). The resulting translation matrix (SVD of E) is wrong when accounting for scale.

I tried both FM_RANSAC and FM_LMEDS. The second one one gives better results in some cases, but in most they are both wrong. Is there some kind of restriction on the points that are given to the function (i.e. they should not be coplanar, which the checkerboard corners are)?

Using x=(cv::Mat)(fs["X"]) produces the error "cannot convert from 'cv::FileNode' to 'cv::Mat' "

When trying to save a cv::Matx to a file using FileStorage, openCV throws some errors which say saving a Matx is not yet implemented. A workaround is to cast the Matx to a cv::Mat when writing:

cv::FileStorage fs("file.xml", cv::FileStorage::WRITE);
cv::Matx33d x;
fs<<"X"<<cv::Mat(x); // file contains x

However, when loading a cv::Matx from file, simply casting it to cv::Mat does not work. I assume the result is not copied, and the Matx is left uninitialized:

cv::FileStorage fs("file.xml", cv::FileStorage::READ);
cv::Matx33d x;
fs["X"]>>cv::Mat(x); // x contains gibberish

Is there a simple workaround for this, without resorting to creating temporary cv::Mat's and casting them to Matx after reading?