This is the code for an advanced practical, completed at the Univerity of Heidelberg in 2017.

• SCIP, which in turn needs GMP
• png++
• VLFeat
• OpenCV

If your're a lazy person and happen to be a Nix user, just use the provided shell.nix file.

Clone this repository using

git clone https://github.com/daniiki/image-segmentation-scip.git 

Then you need to dowlnload the scipoptsuite from http://scip.zib.de/#download and untar the archive. Make sure to place the scipoptsuite-4.x.x directory inside the root folder of this project. Finally, you can compile SCIP:

cd scipoptsuite-4.0.0 make ZIMPL=false READLINE=false 

and then our program:

cd .. make ZIMPL=false READLINE=false 

Finally, you can run it:

bin/fopra input.png 20 

where 20 is the desired number of superpixels.

Have a look at https://daniiki.github.io/image-segmentation-scip. There are also slides about this project at https://github.com/daniiki/image-segmentation-scip/blob/master/presentation/slides.pdf.