Abstract

Identifying spatial regions where biodiversity is threatened is crucial foreffective ecosystem conservation and monitoring. In this stydy, we assessedvarios machine learning methods to detect grazing trails automatically. Wetested five semantic segmentation models combined with 14 different encodernetworks. The best combination was UNet with MambaOut encoder. The solutionproposed could be used as the basis for tools aiming at mapping and trackingchanges in grazing trails on a continuous temporal basis.

 

Quick Read (beta)