[GRASS-user] Using i.smap for segmentation only possible?

[Nikos Alexandris]

On Thu, 2008-07-31 at 14:18 +0200, Moritz Lennert wrote:
> On 31/07/08 12:47, Markus Neteler wrote:
> > On Thu, Jul 31, 2008 at 11:39 AM, Nikos Alexandris
> > <nikos.alexandris at felis.uni-freiburg.de> wrote:
> >> How could I use i.smap to perform segmentation on an image (e.g. aerial
> >> photo) and get the segment only? Unfortunately I don't have now the
> >> GRASS-book with me in which, I remember, there was a reference about
> >> that.
> > 
> > Yes.
> > Just create a group/subgroup with one raster map in. Then
> > go ahead as usual.
> 
> But the way i.smap is conceived now, you have to give it training maps, 
> and so you define the number of types it creates. Doesn't this influence 
> the segmentation results ? Is it possible to just ask for the 
> segmentation without the classification ?
> 
> Moritz

Thank you Markus.

My 2nd concern was exactly what Moritz describes above. But without
being sure that the 1st works I didn't dare to ask further. Probably the
reply is already in the 40something pages that presents the SMAP
algorithm (look at the i.smap manual). Anyhow, it would be nice to have
a one-word answer as well from someone who understands i.smap

Even further: in one commercial segmentation package I' ve used in the
past (e-cognition), one can influence the size of the segments somehow
(nobody knows how exactly except of the authors of the package). How
would we control the size of segments using i.smap if possible?