Table 4 Ablation experiment with decoder.

From: Research on the performance of the SegFormer model with fusion of edge feature extraction for metal corrosion detection

Model

mIoU \(\uparrow\)

Acc \(\uparrow\)

FPS \(\uparrow\)

Params \(\downarrow\)

Flops \(\downarrow\)

SegFormer

66.20

85.32

66.37

3.72

6.78

SegFormer-LawinASPP49

59.33

81.36

34.95

3.45

3.09

SegFormer-FFM

66.25

86.03

69.81

3.58

2.79

  1. Significant values are in bold.
Back to article page