Solar panel counting using SAM

[shamukshi]

For "solar panel counting from UAV image using SAM"....
which approach is better >>> 1. 'objection detection followed by segmentation' or 2. 'segmentation followed by classification'

[giswqs]

See this example: https://samgeo.gishub.org/examples/swimming_pools/

[jamescrowley]

@shamukshi did you have any success with this? I've been experimenting with the model using the swimming pool example as a base, but not found an effective combination of parameters and prompts yet - most of the results are way off.

[shamukshi]

hi... so my target was solar panel.....so when i use "solar panels" as prompt...it gave me solar panel... When I used "glass" as a prompt....it still gave me solar panels... a bit same for other prompts too... I guess that AI prompt technology is still in the works. So I just went with the traditional method...I used SAM to create segments (unclassified segments) from my UAV images... Trained a classifier to classify those SAM segments as solar panel and non-solar panel

…

On Wed, 11 Sept 2024 at 13:47, James Crowley ***@***.***> wrote: @shamukshi <https://github.com/shamukshi> did you have any success with this? I've been experimenting with the model using the swimming pool example as a base, but not found an effective combination of parameters and prompts yet - most of the results are way off. — Reply to this email directly, view it on GitHub <#193 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AN3XPLPVHCH5LZMW7NL2QN3ZV74AFAVCNFSM6AAAAABOAMURPSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTANRRGEZDGOA> . You are receiving this because you were mentioned.Message ID: ***@***.*** .com>