Using AAEHS-Net as an Attention-Based Auxiliary Extraction and Hybrid Subsampled Network for Semantic Segmentation

Table 8

Results of segmentation on the Massachusetts roads dataset.

Method	Accuracy	Dice	MIOU	AUC
U-Net	0.9535	0.6356	0.5138	0.9082
FPN	0.9517	0.6370	0.5153	0.9056
Deeplabv3	0.9540	0.6326	0.5207	0.9071
MHSU-Net	0.9577	0.6401	0.5231	0.9146
HAD-ResUNet	0.9596	0.6397	0.5242	0.9140
MR-UNet	0.9563	0.6441	0.5228	0.9149
CE-UNet	0.9571	0.6437	0.5230	0.9157
Ours	0.9623	0.6470	0.5265	0.9171

The bold value indicates the best value for the item.