i.segment vs. i.segment.gsoc

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

i.segment vs. i.segment.gsoc

mebrau_UT
Hello list,
I am working on segmentation of aerial imagery in forest areas and have
used both modules (i.segment and i.segment.gsoc) for some experiments.
When running i.segment and i.segment.gsoc with radioweight = 1 on the
same dataset with a threshold of 0.3, I assumed the output should be
similar, but it isn't.
In the manual of i.segment.gsoc it is stated that the threshold will be
multiplied by the number of rasters in the image group, so I tried a
threshold of 0.1 in i.segment.gsoc vs a threshold of 0.3 in i.segment as
I have 3 rasters in my image group. Although the results are now more
similar, they still vary a lot.
I have attached the images to show the derived segments from i.segment
0.3 and i.segment.gsoc 0.1. Can someone help me to understand why this
is happening? I would appreciate it.
Kind regards,
Melanie Brauchler
_______________________________________________
grass-user mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/grass-user

segments.png (503K) Download Attachment
segments_gsoc.png (477K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: i.segment vs. i.segment.gsoc

Moritz Lennert
i.segment.gsoc is the original version developed during GSoC. i.segment is ann optimized version that was originally based on the GSoC work, but has known different developments since, including in the way the threshold parameter is applied.

I would recommend using i.segment.

Moritz

Am 10. Juni 2020 13:13:33 MESZ schrieb [hidden email]:

>Hello list,
>I am working on segmentation of aerial imagery in forest areas and have
>
>used both modules (i.segment and i.segment.gsoc) for some experiments.
>When running i.segment and i.segment.gsoc with radioweight = 1 on the
>same dataset with a threshold of 0.3, I assumed the output should be
>similar, but it isn't.
>In the manual of i.segment.gsoc it is stated that the threshold will be
>
>multiplied by the number of rasters in the image group, so I tried a
>threshold of 0.1 in i.segment.gsoc vs a threshold of 0.3 in i.segment
>as
>I have 3 rasters in my image group. Although the results are now more
>similar, they still vary a lot.
>I have attached the images to show the derived segments from i.segment
>0.3 and i.segment.gsoc 0.1. Can someone help me to understand why this
>is happening? I would appreciate it.
>Kind regards,
>Melanie Brauchler
_______________________________________________
grass-user mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: i.segment vs. i.segment.gsoc

mebrau_UT
Thank you for the explanation! I assumed gsoc version is build upon i.segment.
Is there any way to include the shape parameters in i.segment? I am very fond of the gsoc version as it helps surpress segments growing to very complex geometries (shadows through the forest patches for example). 

Kind regards,
Melanie

On Jun 11, 2020 12:52 PM, Moritz Lennert <[hidden email]> wrote:

i.segment.gsoc is the original version developed during GSoC. i.segment is ann optimized version that was originally based on the GSoC work, but has known different developments since, including in the way the threshold parameter is applied.

I would recommend using i.segment.

Moritz

Am 10. Juni 2020 13:13:33 MESZ schrieb [hidden email]:
>Hello list,
>I am working on segmentation of aerial imagery in forest areas and have
>
>used both modules (i.segment and i.segment.gsoc) for some experiments.
>When running i.segment and i.segment.gsoc with radioweight = 1 on the
>same dataset with a threshold of 0.3, I assumed the output should be
>similar, but it isn't.
>In the manual of i.segment.gsoc it is stated that the threshold will be
>
>multiplied by the number of rasters in the image group, so I tried a
>threshold of 0.1 in i.segment.gsoc vs a threshold of 0.3 in i.segment
>as
>I have 3 rasters in my image group. Although the results are now more
>similar, they still vary a lot.
>I have attached the images to show the derived segments from i.segment
>0.3 and i.segment.gsoc 0.1. Can someone help me to understand why this
>is happening? I would appreciate it.
>Kind regards,
>Melanie Brauchler



_______________________________________________
grass-user mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/grass-user
Reply | Threaded
Open this post in threaded view
|

Re: i.segment vs. i.segment.gsoc

Moritz Lennert
On 11/06/20 13:36, [hidden email] wrote:
> Thank you for the explanation! I assumed gsoc version is build upon
> i.segment.
> Is there any way to include the shape parameters in i.segment? I am very
> fond of the gsoc version as it helps surpress segments growing to very
> complex geometries (shadows through the forest patches for example).

There was some reasons why the shape parameters were not integrated in
i.segment, but I can't remember why. Some code is actually there, but
currently it is not used.

MarkusM, do you remember ? And how complicated would it be to integrate
them ?

Moritz


>
> Kind regards,
> Melanie
>
> On Jun 11, 2020 12:52 PM, Moritz Lennert <[hidden email]>
> wrote:
>
>     i.segment.gsoc is the original version developed during GSoC.
>     i.segment is ann optimized version that was originally based on the
>     GSoC work, but has known different developments since, including in
>     the way the threshold parameter is applied.
>
>     I would recommend using i.segment.
>
>     Moritz
>
>     Am 10. Juni 2020 13:13:33 MESZ schrieb [hidden email]:
>      >Hello list,
>      >I am working on segmentation of aerial imagery in forest areas and
>     have
>      >
>      >used both modules (i.segment and i.segment.gsoc) for some
>     experiments.
>      >When running i.segment and i.segment.gsoc with radioweight = 1 on the
>      >same dataset with a threshold of 0.3, I assumed the output should be
>      >similar, but it isn't.
>      >In the manual of i.segment.gsoc it is stated that the threshold
>     will be
>      >
>      >multiplied by the number of rasters in the image group, so I tried a
>      >threshold of 0.1 in i.segment.gsoc vs a threshold of 0.3 in i.segment
>      >as
>      >I have 3 rasters in my image group. Although the results are now more
>      >similar, they still vary a lot.
>      >I have attached the images to show the derived segments from
>     i.segment
>      >0.3 and i.segment.gsoc 0.1. Can someone help me to understand why
>     this
>      >is happening? I would appreciate it.
>      >Kind regards,
>      >Melanie Brauchler
>
>
>
> _______________________________________________
> grass-user mailing list
> [hidden email]
> https://lists.osgeo.org/mailman/listinfo/grass-user
>

_______________________________________________
grass-user mailing list
[hidden email]
https://lists.osgeo.org/mailman/listinfo/grass-user