Jian Kang, Ruben Fernandez-Beltran, Xian Sun, Jingen Ni, Antonio Plaza
This repo contains the codes for the GRSL paper: Deep Learning-based Building Footprint Extraction with Missing Annotations. We propose a novel loss function for extracting building footprints based on the training images with missing annotations. The loss includes three terms: 1) logit adjusted cross entropy (LACE) loss, aimed at discriminating between building and background pixels from a long-tailed label distribution; 2) weighted dice loss, aimed at increasing the F1 scores of the predicted building masks; and 3) boundary alignment loss, which is optimized for preserving the fine-grained structure of building boundaries.
train/main_.py is the script of the proposed method for training and validation.
utils/loss_.py contains the considered losses for building footprint extraction.
Some codes are modified from SegLoss and Geoseg.
@article{kang2021bdsegma,
title={{Deep Learning-based Building Footprint Extraction with Missing Annotations}},
author={Kang, Jian and Fernandez-Beltran, Ruben and Sun, Xian and Ni, Jingen and Plaza, Antonio},
journal={IEEE Geoscience and Remote Sensing Letters},
year={2021},
note={DOI:10.1109/LGRS.2021.3072589}
publisher={IEEE}
}