amueller / gco_python

Python wrappers for GCO alpha-expansion and alpha-beta-swaps
137 stars 65 forks source link

non-symmetric smooth_cost #2

Closed fabianp closed 12 years ago

fabianp commented 12 years ago

Hey!

Do you know if unsymmetric smoothness_cost is supported?. The Matlab interface (function GCO_SetSmoothCost) checks for that and raises an error in that case. Some sentences in the GCO_README also suggest that it's not supported.

If that is the case, I suggest adding the check, it will save headaches to future users :-)

amueller commented 12 years ago

Hey Fabian. Thanks for noticing. I think they are not supported. I'll add a check when I find the time (...). If you want that feature, try qpbo, that can also handle non-submodular terms.

amueller commented 12 years ago

Btw, would you join foces to write a cython Boykov-Kolmogorov implementation? That is my next project after segmentation for skimage. Are you working on grid graphs?

fabianp commented 12 years ago

Thanks for the link, that might be just what I need!

I'll think about writing a cython Boykov-Kolmogorov, but for now I'm just trying a few things and don't have a clear direction. In the meantime I'll send you a couple more of patches for this :-)

amueller commented 12 years ago

Thanks, all patches welcome! I have been reading some inference papers recently, so if you want to discuss, I'm idling at scikit-learn irc.

amueller commented 12 years ago

btw, many people seem to use TRW-S. Have you heard of that? I haven't written any wrappers for that, yet.

amueller commented 12 years ago

Should be fixed in ef75b24fe720b4b4d88ecbddb6695d85df02ad98.