ysalaun/mlsd

mlsd's Issues

windows issue

Great job, I use the MLSD in Windows 10, but I find some bugs and runtime too long !

Can I use this with the LBD descriptor?

I would like to use your MLSD line feature to do visual odometry.

So can I do line matching using MLSD line feature and the LBD descriptor?

Thank you!

python wrapper

Hello

Thank you for this algo.
It will be very nice to get a python wrapper.

Windows compatibility

Computation time too long
Bad results with test images

Why my code testing result is incorrect?

Hi, ysalau ,
I'm very interested in your research. I try to test your code on picture Office.png as follow :
//-------------------------------------my test code---------------------------------------------------
#include "detection.hpp"
using namespace cv;
int main(void)
{
Mat img = imread("Office.png"); //read the picture
vector ImagePyramid;
ImagePyramid = computeImagePyramid(img, true);
vector Seg;
Seg = lsd_multiscale(ImagePyramid,0.1,true);
Mat showseg(img.rows,img.cols, CV_8UC1, Scalar(0));
for (int i=0;i<Seg.size();i++)
{
Point p1(Seg[i].x1, Seg[i].y1);
Point p2(Seg[i].x2, Seg[i].y2);
line(showseg, p1, p2, Scalar(255));
cout << "x1 = "<<Seg[i].x1<< " y1 = " << Seg[i].y1<< " x2 = " << Seg[i].x2<< " y2 = " << Seg[i].y2 << endl;
}
imwrite("C:/Users/lzlong/Desktop/mlsd_2/test2/showseg.jpg" , showseg);
return 0;
}
//-------------------------------------------------end code---------------------------------------------

I think the Seg is the result of the test. So , I try to output the result on the picture "showseg.jpg" .But I find the picture "showseg.jpg" only horizontal lines. And then ,I check every Seg[i]'s (x1,x2) and (x2,y2),I find that every segment of output is horizontal lines.
Do I use the code correctly? Why the result would be like this？
I upload the test result:

Ambiguity in README

"This code is part of the following publication and was subject to peer review"
lets believe the source code was peer-reviewed. Only the article was. I propose to reformulate:
"This code is linked of the following publication, which was subject to peer review"

Pascal