Segmenting remote sensing imagery with the Segment Anything Model and box prompts

[@mahsaghandaliiigh9657]

Thank professor, is it doable for tif format too?

[@ndbharti]

That's cool, Thank you professor. Wonder if in a bounding box have number of buildings and trees, what should be the output ?
How many number of polygon will be there?

[@giswqs]

You can have overlapping polygons. One for each type

[@ndbharti]

At the end we are getting output as a polygon. When you zoom in that polygon looks like irregular, how to smooth that polygon and polygon must cover all the vertexes.
If building is in L, K shape shape so our output polygon should be in L or K shape respectively.

[@ndbharti]

@@giswqs today I'll give a try on box prompts as well, and let you know.
Thank you professor

[@znk4076]

This is great! Thank you Professor Wu! I wonder if this can also apply to building roofs?

[@giswqs]

I think so! Give it a try.

[@Md.Ashikuzzaman-q4g]

Hello Professor, Can you please make a video this process using google colab as I am having problem with my jupyter notebook.

[@loganwoolfson5823]

Hi ​ @giswqs when I run the code using your my own image, I get a few errors. Any suggestions? The tif image I am using is around 7 GB and has 6 bands (high resolution UAV orthophoto). Any advice is helpful.

[@aureliopds]

Professor, thanks for your amazing job! I wonder if there is a image size limit,? since i'll try with agriculture images, something around 10000 ha

[@ahmetonurevis6989]

I'm encountering the following error, can you help?
note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for GDAL
Failed to build GDAL
ERROR: Could not build wheels for GDAL, which is required to install pyproject.toml-based projects