[GRASS-dev] [GRASS-web] GRASS i.segment

Vaclav Petras wenzeslaus at gmail.com
Tue May 11 18:51:41 PDT 2021


Hi Guy,

The option #1 is still available to you as far as I can tell. You can
actually use Docker images with AWS Lambda and that was introduced exactly
to solve dependency issues. The images can be in GBs. That's quite
sufficient for GRASS GIS, its dependencies, and any other Python packages
you need. Adding other Python packages which your code needs should not be
a problem.

With that said, #2, i.e., pip or conda installation would be still helpful
in many cases including yours (although there would still be a few hurdles
before putting it on the old container-free AWS Lambda). Same goes for
minimizing the build (like including only components necessary for
i.segment). Seemingly unrelated, but CMake build support (now as a
prototype) may be a step towards better packaging. Anyway, this would be
all optional if you go with #1.

I would also add item #3 which is something I suggested earlier. You did
not mention it explicitly, but it seems to me that your workflow would
benefit from an API oriented on execution of a single module. Although
that's certainly possible, a different API may provide a smoother
experience in these cases. This does not replace #1 or #2, but it is
complementary.

I have cc-ed the grass-dev mailing list which is what the developers are
following more closely. The best way is, however, to subscribe to that list
and post your goal, possibly with some of my points, there with "funding
available" included in the subject.

https://lists.osgeo.org/mailman/listinfo/grass-dev

Best,
Vaclav

On Tue, May 11, 2021 at 9:40 AM Sinter Guy <Guy.Sinter at manna-irrigation.com>
wrote:

> Hi,
>
>
>
> Any response ?
>
> Thanks,
>
> Guy
>
>
>
> *From:* Sinter Guy
> *Sent:* יום ה 29 אפריל 2021 13:04
> *To:* Vaclav Petras
> *Cc:* grass-web at lists.osgeo.org; Beeri Ofer; Pelta Ran; Mor Eyal
> *Subject:* RE: [GRASS-web] GRASS i.segment
>
>
>
> Hi,
>
>
>
> Our  runtime environment is AWS lambda and there are other Python packages
> which need to be installed with our code,
>
> So we cannot use option number #1 below,
>
>
>
> Regarding Option #2 im not sure this is what we thought of, as mentioned
> before I believe  that what we and others (based on my web search for
> solution)
>
> are looking for is a way to execute I segment purely from python without
> an installation, I think this is what you meant in this option,
>
> as we said we are willing to pay for such development which will
> contribute to the entire community ,
>
> let me know if its possible,
>
> Thanks,
>
> Guy
>
>
>
> *From:* Vaclav Petras [mailto:wenzeslaus at gmail.com]
> *Sent:* יום ה 29 אפריל 2021 04:00
> *To:* Sinter Guy
> *Cc:* grass-web at lists.osgeo.org; Beeri Ofer; Pelta Ran; Mor Eyal
> *Subject:* Re: [GRASS-web] GRASS i.segment
>
>
>
> *EMAIL FROM EXTERNAL SENDER: *Do not click links or open attachments
> unless you know that the sender and the content are safe.
>
>
>
> To add to what Moritz said, it sounds like the separate installation is
> the main issue you have. There is one existing solution and one which needs
> to be implemented.
>
>
>
> 1) Installing GRASS GIS is not that big of a deal on many platforms, e.g.,
> on Linux machines or in a Linux Docker container. The difference is
> installing a system package versus a Python package. That works well for
> many because GRASS GIS installation is considered small by most people.
> Sure there are dependencies, some things you don't need, etc., but when one
> is using Python and is not on an embedded system, it is fine.
>
>
>
> 2) Your installation trouble might not exist if GRASS GIS is installable
> alongside other Python packages with pip or conda. That's not possible
> right now, but not impossible. A conda environment is used to build GRASS
> GIS for macOS and there was a prototype conda package, too. Since I tested
> these myself, I can say that's definitely some time/money investment - but
> a good one.
>
>
>
> Let us know what you think.
>
>
>
> Best,
>
> Vaclav
>
>
>
> On Wed, Apr 28, 2021 at 1:36 AM Sinter Guy <
> Guy.Sinter at manna-irrigation.com> wrote:
>
> Hi Vaclav,
>
>
>
> Thank you for replying ,
>
> We are looking for a standalone python package of the i.segment that could
> run as part of independent code without the need for the GRAA installation,
>
> Is this something that is on your road map ?
>
> Is this a feature we can support by paying for  ?
>
>
>
> Thanks,
>
> Guy
>
>
>
> *From:* Vaclav Petras [mailto:wenzeslaus at gmail.com]
> *Sent:* יום ד 28 אפריל 2021 04:20
> *To:* Sinter Guy
> *Cc:* grass-web at lists.osgeo.org; Beeri Ofer; Pelta Ran
> *Subject:* Re: [GRASS-web] GRASS i.segment
>
>
>
> *EMAIL FROM EXTERNAL SENDER: *Do not click links or open attachments
> unless you know that the sender and the content are safe.
>
>
>
> Hi Guy,
>
>
>
> I don't know if you got a private answer, but the general answer from the
> GRASS project is that we are interested in making the algorithms as
> available as possible as long as they access the canonical implementation
> in GRASS GIS.
>
>
>
> In other words, if you aim at a better interface in Python to i.segment,
> that is something which aligns well with the goals of the GRASS project.
> The current Python interface is fine, but not the best imaginable. Same
> goes for installation of GRASS GIS. It is fine, but there could be, for
> example, more options. So, it depends what routes you are willing to take.
>
>
>
> Please note that I'm a Project Steering Committee member and while trying
> to represent the project here as best as I can, I'm ultimately speaking
> from my point of view.
>
>
>
> In any case, thank you for bringing this idea and offer to the community.
>
>
>
> Best,
>
> Vaclav
>
>
>
>
>
> On Mon, Apr 19, 2021 at 7:59 AM Sinter Guy <
> Guy.Sinter at manna-irrigation.com> wrote:
>
> Hi Grass GIS team
>
>
>
> My name is Guy and I’m from Manna Irrigation we provide farmers with
> irrigation recommendation and fields images,
>
> We are working on an algorithm to identify  point of interest within field
> and for that our remote sensing team have used  your i.segment as part of
> their algorithm,
>
> We would like to ask if you are planning to develop the i.segment package
> as a standalone python package without the need for Grass GIS installation ?
>
> If so can you share the details ? we are also willing to pay for such
> development,
>
>
>
> Thanks Much ,
>
> *Guy Sinter*
>
> *Director, Software Engineering*
>
> +972 52 6433180 (Cell)
>
> www.manna-irrigation.com
>
> [image: manna]
>
>
>
> _______________________________________________
> grass-web mailing list
> grass-web at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/grass-web
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20210511/2f56c29b/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 18782 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20210511/2f56c29b/attachment-0001.jpg>


More information about the grass-dev mailing list