A SAM-based unsupervised change detection method that matches and splits segmentation masks across two dates, improving F1 over AnyChange on GZ_CD_data.
Segment anything,
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
-
MergeSAM: Unsupervised change detection of remote sensing images based on the Segment Anything Model
A SAM-based unsupervised change detection method that matches and splits segmentation masks across two dates, improving F1 over AnyChange on GZ_CD_data.