SST-CD trains an end-to-end building change detector from unlabeled bi-temporal images by using temporal discrepancies as pseudo labels filtered by spatial consistency, achieving F1 scores of 83.08%, 91.69%, and 86.60% on three datasets.
Unsuper- vised universal image segmentation,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Spatially Selective Self-Training for Unsupervised Building Change Detection
SST-CD trains an end-to-end building change detector from unlabeled bi-temporal images by using temporal discrepancies as pseudo labels filtered by spatial consistency, achieving F1 scores of 83.08%, 91.69%, and 86.60% on three datasets.