A good architecture for satellite image segmentation?

I’m trying to perform an image segmentation task using satellite images. I’ll be using the unet_learner. But which architecture might give me good results?

I’m ready to experiment with architectures myself but I want to know what others have done and the problems they faced (and if they solved it, how?).

Links to papers are also welcome.

Thank you :smiley:

I tried to implement the same using segnet architecture and got some decent results long ago. I have taken the pipeline from here. It was written end-to-end in pytorch. It might be helpful for you!

1 Like