Semantic Segmentation Algorithm Based on Attention Mechanism and Transfer Learning

Table 5

The comparison of the performance of semantic segmentation models based on transfer learning.

Dataset	Model	Before the migration MIoU (%)	After the migration MIoU (%)
Cityscapes	FCN wild	56.20	60.50
Migrated dataset	Curriculum DA	56.30	60.80
	TransRoadNet	58.60	62.70
Surv-Cityscapes	FCN wild	5.90	24.30
Migrated dataset	Curriculum DA	6.10	26.90
	TransRoadNet	6.20	30.60
Virt-Cityscapes	FCN wild	21.10	27.10
Migrated dataset	Curriculum DA	22.30	31.40
	TransRoadNet	23.10	35.80