Negative number of test results

[wangshuo98]

Hi ,
Thanks for you great m-lsd work!
I use it to detect my own pictures,and save the Endpoint coordinates.There are small amount of Endpoint coordinates is Negative number .I find the mistake only happened at x_start.Can you give me some information of it .Thanks a lot
a result file
1403638127445096960.txt

[geonm]

Hi.

I guess it's due to a wrongly predicted center point.

A displacement vector for the center point is predicted by reasoning both ends of a line segment relative from the center point, but if the predicted center is located far from one side and the other endpoint is out of an image, it tries to find a virtual point outside the image.

Can you share an example with me?

[wangshuo98]

Thank you for your prompt reply before the Lunar New Year！

I'd be happy to provide pictures for you to test . I use mlsd on EuRoC dataset （MH03 MH04） score_thr=0.4, dist_thr=10;

If you have an idea to solve this problem, please let me know.Thanks a lot.

[geonm]

Happy lunar new year too!

I've tested your examples and found it as I guessed before.

You can easily solve this problem.

If you'd like to fix a predicted line segment outside of an input image, just find an intersection between the predicted line segment and an edge of the image in the image coordinate system.

Note the edge of the image is one of these lines; (0,0)~(0,h), (0,h)~(h,w), (h,w)~(0,w), (0,w)~(0,0)

You can refer to https://github.com/navervision/mlsd/blob/master/utils.py#L176-L195 for this.