[GRASS-dev] how difficult would it be to modify i.edge to not load the entire image into memory ?

[Markus Metz]

On Thu, Dec 14, 2017 at 12:44 PM, Markus Metz <markus.metz.giswork at gmail.com>
wrote:
>
>
>
> On Thu, Dec 14, 2017 at 11:47 AM, Moritz Lennert <
mlennert at club.worldonline.be> wrote:
> >
> > On 14/12/17 05:45, Vaclav Petras wrote:
> >>
> >> Hi Moritz,
> >>
> >> I'm currently at AGU
> >
> >
> > Enjoy ! :-)
> >
> >> and can't dig into the code, but yes, I think it can be rewritten with
segment library, preferably with MarkusM's "segment or in-memory" wrapper.
> >
> >
> > Great ! Markus, is this wrapper in the library, or just in existing
modules ? I know the version in the i.superpixels.slic addon and have some
memory of you implementing it elsewhere.
>
> I want to put it in the library, but would like to simplify the cache a
bit before putting it in the library. Be aware that addons need their local
version of this cache, otherwise they won't compile against previous
versions of GRASS 7.

I have simplified the cache for i.superpixels.slic in r71936. The
all-in-memory mode uses now less memory and is faster :-)

Markus M
>
> Markus M
>
> >
> > Moritz
> >
> >
> >
> >>
> >> Vaclav
> >>
> >> On Tue, Dec 12, 2017 at 10:51 AM, Moritz Lennert <
mlennert at club.worldonline.be <mailto:mlennert at club.worldonline.be>> wrote:
> >>
> >>     Hi Anna, Hi Vaclav,
> >>
> >>     I would love to use the i.edge module. However, the fact that it
> >>     reads everything into memory obviously limits its use either to
> >>     small images, or to tiled versions of larger images.
> >>
> >>     I have just rapidly looked at the code, but I cannot really detect
> >>     any place where you really need the entire image, or ? The Gaussian
> >>     filter could be cut off at a certain distance and the actual Canny
> >>     detector is also based on a kernel, AFAICT. So, do you think this
> >>     could be rewritten to use the SEGMENT library and at all times only
> >>     load necessary parts of the image into memory ?
> >>
> >>     Moritz
> >>
> >>
> >
> > _______________________________________________
> > grass-dev mailing list
> > grass-dev at lists.osgeo.org
> > https://lists.osgeo.org/mailman/listinfo/grass-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20171214/1b354643/attachment.html>