IGraSS iteratively refines noisy canal maps by enforcing reachability to water sources, reducing unreachable canal pixels from about 18% to 3% and improving segmentation in held-out test patches.
Joint topology-preserving and feature-refinement network for curvilinear structure segmentation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
IGraSS: Learning to Identify Infrastructure Networks from Satellite Imagery by Iterative Graph-constrained Semantic Segmentation
IGraSS iteratively refines noisy canal maps by enforcing reachability to water sources, reducing unreachable canal pixels from about 18% to 3% and improving segmentation in held-out test patches.