Integration of SEEDS superpixels algorithm

asked 2014-07-03 11:21:04 -0600

bkueng
56 ●1 ●4

I'd like to include a new image segmentation algorithm in OpenCV, but I first want to make sure whether it would be accepted at all (with the necessary coding style adaptations of course)?

It is the SEEDS superpixels algorithm, described here (with source code). It is currently one of the fastest (probably even the fastest) high-accuracy oversegmentation algorithm.

If the algorithm is acceptable, is it ok to use C++ classes? And if so, can it then be used with python?

Thanks

edit retag flag offensive close merge delete

add a comment

2 answers

Sort by » oldest newest most voted

answered 2014-07-03 12:35:43 -0600

Mathieu Barnachon

4678 ●18 ●53 http://www.math-barnac...

I think every contribution is interesting! See this page on how to contribute if you haven't. It will explain the pull request method used in OpenCV. What is important, to my point of view:

be sure to have the right of the code you are using (or specify it if not free)
create a code with the OpenCV standard
don't forget the doc!
the binding in other languages (like Python) could be difficult, but it's worth it!

Good luck, I hope this will be soon in OpenCV... ;-)

edit flag offensive delete link

Comments

Good exeplanation! Maybe @berak can comment on the wrappers, guess at least the python-wrappers are not too difficult to create...

Guanta ( 2014-07-03 14:12:24 -0600 )edit

i don't think the python/java wrappers are any difficult, once you stick to c++ and and a similar way the other opencv algos are handling args (like use Input/OutputArrays, references, avoid pointers)

usually adding support for the wrappers only means adding CV_EXPORTS_W to the class declaration, and adding CV_WRAP to any method you want exposed.

berak ( 2014-07-03 14:24:14 -0600 )edit

hi, has this been implemented in version 3 (beta)? i saw in the online docs that there is now a set of functions for SuperpixelSEEDs but I can't find them in the build. thanks