Table 4.

Comparison of segmentation models.

Model	Pyramid pooling	Edge detection focus	Boundary precision	Computational efficiency	Key strengths
UNet	✗	✗	Moderate	High	Strong performance in small datasets
SegNet	✗	✗	Moderate	Moderate	Memory-efficient decoding
PSPNet	✓	✗	Moderate to High	Moderate	Multiscale context understanding
DeepLabv3	✓	✗	Moderate to High	Moderate to Low	Captures large context without resolution loss
Swin-Unet	✓	✗	High	High	Good for dense prediction tasks
Light weight transformer model based on ViT	✗	✗	Moderate	Low	Excellent for classification and transfer learning
PSPEdgeWeedNet	✓	✓	High	Moderate	Improved boundary localization, efficient deep feature extraction, better fine-grained segmentation