Table 1 Evaluation metrics results of ablation experiments on the Potsdam dataset with different configurations of the network.

From: A deep learning method for optimizing semantic segmentation accuracy of remote sensing images based on improved UNet

Model

CACB

AFAM

SAM

OA

mIOU

Unet

   

82.01

66.44

UNet + CACB

  

84.46

69.96

UNet + CACB + AFAM

 

85.13

71.24

AFF-UNet

85.4

71.44

  1. OA and mIOU are both expressed in percentages.