Table 19 Ablation study performance of DAC-GAN.

From: Deep atrous context convolution generative adversarial network with corner key point extracted feature for nuts classification

CNN Based Models

Input type

Acc (%)

Mean (%)

Var

STD (%)

Pre (%)

Recall (%)

IoU (%)

(%) F-Score

Baseline GAN with raw images

RGB

92.76

92.89

0.52

0.72

92.58

92.63

90.89

92.60

GAN with CKPF images

RGB

94.83

94.95

0.41

0.64

94.78

94.81

93.02

94.80

GAN with default atrous convolution (raw)

Grayscale

96.47

96.55

0.32

0.57

96.61

96.44

94.72

96.53

GAN with atrous + pre-context block (raw)

Grayscale

97.24

97.31

0.27

0.52

97.33

97.25

95.89

97.29

GAN with atrous + post-context block (raw)

Grayscale

97.48

97.54

0.25

0.50

97.51

97.46

96.08

97.43

GAN with default atrous convolution (CKPF)

Grayscale

98.40

98.47

0.19

0.44

98.41

98.29

96.98

98.35

GAN with atrous + pre-context block (CKPF)

Grayscale

99.33

99.39

0.14

0.37

99.28

99.24

98.53

99.27

GAN with atrous + post-context block (CKPF)

Grayscale

99.52

99.57

0.09

0.30

99.44

99.47

99.08

99.45

DAC-GAN with CKPF (RGB)

RGB

98.56

98.61

0.07

0.26

98.50

98.53

98.41

98.52

Proposed DAC-GAN with CKPF (Grayscale)

Grayscale

99.83

99.89

0.04

0.20

99.85

99.89

99.69

99.69