Bag of Contour Fragments - Experience/Source Code
Not sure if this is the proper forum for asking but here goes:
I am working on trying to match partially occluded shape contours and am trying various approaches. Most recently, I tried Generalized Hough Transform and that may be my fallback position. That said, I am looking for something that gives better recognition in shorter times.
I am considering implementing the contour fragment matching approach described in: Xingang, et al., "Bag of Contour Fragments for Robust Shape Classification" (2014). http://dl.acm.org/citation.cfm?id=258...
Does anyone have any experience with this approach that they would be willing to share?
More importantly, has anyone implemented this using OpenCV and would they be willing to share some code?
This method relies on Discrete Contour Evolution and code for that would also be very helpful.
For those interested, Matlab code is available (https://bitbucket.org/xinggangw/bcf/src) but I am hoping to avoid porting it.
Thanks, James
I was looking for an algorithm which detects an texture-less object, too. My problem is, that an algorithm which uses a shape template (for example a dxf-shape as edge map) doesn't seem to be available.
The most algorithms need a grayscale image to get a gradient magnitude and/or direction such as GeneralizedHough. The only exception seems to be chamferMatching which was removed in OpenCV 3.
If you have a color template you can take a look at linemod. I think this could also be a good base to enhance this algorithm for grayscale templates or even shape templates.
"exception seems to be chamferMatching which was removed in OpenCV 3" if somebody make a PR with this code I think it can be opencv 3