I am using the module i.segment.uspo in GRASS for the first time and I wanted to know if anyone here haws experience with that add-on in GRASS so I can get some hints on how to use it more efficiently, because I could use i.segment but I feel like it wouldn't be the same because, for once, I don't know how to properly set up meanshift. So, can anyone help me understand better how i.segment.uspo works or are you as new to this as I am?

P.S If someone could point me to some examples, or even share, other than the ones that are in the documentation file, that would be great.


At first set you region of interest with this command:

g.region -d vector=myregion@name_of_grasssession save=myregion1

This command creates region of interest bordered with vector layer placed in grass database called myregion1

In the next step you are able to run optimization process of segmentation:

i.segment.uspo group=myimage@name_of_grass_session regions=myregion1@name_of_grass output=parameters.csv segment_map=ortho_uspo threshold_start=0.02 threshold_stop=0.21 threshold_step=0.02 minsizes=5,10,15 number_best=5 processes=4 memory=4000

It runs pretty well because it is optimized for parallel processing. After algorithm is complete you will be able to find .csv file if you run windows:


I plan to use this tool in my Ph.D. thesis and I have started using this tool so far just like you. If I discover something useful I will share my experience.

  • I am writing a script in shell with i.segment.uspo for my undergrad thesis. But I am stuck at a loop and I can't seem to export a step-based segment-map – George Nostradamos Sep 14 '17 at 7:13

Other than the documentation and this article (https://www.researchgate.net/publication/315897308_An_Open-Source_Semi-Automated_Processing_Chain_for_Urban_Object-Based_Classification) there's nothing else I could find.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.