This work was originally developed by Miguel Gonçalves in the context of an M.Sc. thesis project at Instituto Superior Técnico of the University of Lisbon. The project aimed to further explore the use of general-purpose Vision Transformer models (i.e., SegForme or the Lawin Transformer) for semantic segmentation of remote sensing imagery, comparing the obtained results with current domain-specific approaches (e.g., DC-Swin, FT-UNetFormer, or RemoteNet).
The source code provided in this repository alows one to reproduce the results described on one publication currently under review.
@inproceedings{Goncalves2023,
author = {Gonçalves, Miguel and Martins, Bruno and Estima, Jacinto},
booktitle = {Proceedings of the 6th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery},
title = {A Detailed Analysis on the Use of General-purpose Vision Transformers for Remote Sensing Image Segmentation},
year = {2023}
}