Fig. 8 | Scientific Reports

Fig. 8

From: Dual-stream hybrid architecture with adaptive multi-scale boundary-aware mechanisms for robust urban change detection in smart cities

Fig. 8

Visual comparison of change detection results on the LEVIR-MCI66: Left side (column a) Pre-change images, (column b) Post-change images, (column c) Ground truth masks, (column d) Predictions using single-stream modified PVT-V2 encoder backbone, and right side (column e) Predictions using proposed DSHA model with dual-stream encoder. The first and second rows (columns d and e) represent building change predictions in red color masks, while rows 3 and 4 (columns d and e) show road changes in yellow and building change predictions.

Back to article page