Table 3 Performance comparisons with baselines on the YouTube-w-ALI dataset, unit: %.

From: Transferring enhanced material knowledge via image quality enhancement and feature distillation for pavement condition identification

Model

Acc.

Rec.

Spec.

CNN-3

90.08

87.36

87.03

CNN-2

90.72

86.29

86.47

CNN-1

90.89

89.34

89.65

SqueezeNet

89.14

85.65

85.36

VGG16

90.65

88.46

88.21

VGG19

90.17

89.17

89.01

ResNet50

92.54

90.34

90.35

ResNeXt50

92.78

92.87

96.48

EfficientNet-B3

92.74

92.88

96.42

FocalNet

93.22

93.19

96.65

DenseNet121

94.08

93.27

93.36

MBFN

97.28

97.06

97.36

IQEFD (Our)

98.04

98.01

98.06

  1. The best value of each metric is bold and italic.
Back to article page