Tree Mapping with Segment-Geospaital and Segment Anything Model 2 (SAM 2)

[@aadesh_kale]

This is quite the breakthrough in gis, I'm still learning about the basics but seeing glimpse about more advanced stuff makes me think the ceiling of geospatial is higher than ever. This motivates me to learn more! Thank you for providing such informative tutorials sir 😊

[@mandata143]

Thank you, Professor.

[@KeyserTheRedBeard]

Impressive video, Open Geospatial Solutions. Looking forward to seeing your next upload from you. I tapped the thumbs up icon on your content. Keep up the fantastic work! Your explanation of tree mapping and the use of bounding boxes was very clear. How do you see the integration of real-time data impacting the accuracy of segmentation in future applications?

[@avtomata47]

Thank you for this, Professor! I do have one question tho... Most of these examples use RGB satellite imagery with very high spatial resolution. With this, is it possible to use SAM 2 to produce land cover maps for multispectral satellite imagery such as those taken by Sentinel-2 (10m)?

[@divina_8770]

Yes you can use but it performs worse.

[@jovin.thazhal]

how do u get the bounding boxes manually? pls reply

[@velikilincarslan2264]

How do we count trees?

[@jovin.thazhal]

how can i do automatic segmentation without doing manual segmentation pls replay sir

[@柳永-z8p]

教授，能不用出一期分割岩石和岩石上的裂纹教程

[@xinyakuang864]

I would like to politely ask about segment geospatial installation and jupyterlab. my GPU doesn't seem to support running this function. I used the Aconda Prompt command to create a virtual environment in the local Conda, and then installed segment geospatial in the virtual environment. After the installation was successful, I started jupyterlab in it, but finally returned to me the gpu error. My laptop only has one integrated graphics card. So I want to ask if it is necessary to have an GPU larger than 8G to run this function? 😢

[@johncook1426]

I keep trying to follow along but it is telling me that SamGeo2 is not a valid import from samgeo. Tried installing/upgrading. Any thoughts?

[@giswqs]

Try the notebook on Colab

[@johncook1426]

@@giswqs figured out my version of python was too old of a version.

[@jpereira_geospatial-o6e]

hello! how can i use this using already segmented annotation data in a vector file format? i want to use those annotations for training data and then predict for the entire raster image. Thank you!

[@giswqs]

Yes. See the notebook example

[@jpereira_geospatial-o6e]

@@giswqs i tried but i didnt gey any results (only a black mask image). I dont have the bounding boxes, but the polygons of the mannual annotated segmentation. should i use any other parameter? EDIT: Ok now it worked, i converted EPSG from my bboxes and tiff to 3857. However, the segmentation was only done on the bounding box inputs, isnt there a way to predict for every tree (without bounding box) a segmentation? Because it is not scalable to draw bounding box for a tiff of 300mbs with lots of trees

[@faarehop3967]

​@@jpereira_geospatial-o6e I am facing the same situation. Working on tiff file covering large area.. Did you find a solution to the problem you were facing?

[@jpereira_geospatial-o6e]

@@faarehop3967 unfortunately not really, maybe there is something in the notebook that does that and Im missing that

[@emanuelv5934]

🤯

[@jovin.thazhal]

sir can u provide me crack segmentation for roads,buildings and dams tutorials pls