Table 6.

Ablation study showing the impact of boundary-aware loss and edge detection branch on PSPNet-based segmentation model performance (Significant values are highlighted in bold).

Model variant	Accuracy	Precision	Recall	F1 Score	IoU
PSPNet + Class weights + Boundary Aware Loss	0.8943	0.6945	0.7236	0.7068	0.5875
PSPNet + Class weights + Edge Detection Branch	0.9279	0.7800	0.8003	0.7897	0.6800
PSPNet + Class weights + Boundary Aware Loss +  Edge Detection Branch (Proposed model)	0.9335	0.9399	0.9335	0.9358	0.8797