Table 26 Architecture variant performance comparison.

From: Enhanced paddy leaf disease detection using novel dual metaheuristic loss functions in generative adversarial networks with identity block preservation for thermal image augmentation

Architecture variant

Generator type

Discriminator type

PSNR (dB)

Accuracy (%)

Training time (hrs)

Standard U-Net

U-Net

PatchGAN

28.45 ± 0.74

91.23 ± 1.12

12.8

ResNet generator

ResNet-based

PatchGAN

29.12 ± 0.68

92.67 ± 0.98

15.2

Dense connections

DenseNet-based

PatchGAN

29.78 ± 0.71

93.45 ± 0.89

16.7

Attention mechanism

U-Net + Attention

PatchGAN

30.23 ± 0.65

94.12 ± 0.82

18.4

Multi-scale disc

U-Net

Multi-scale

30.67 ± 0.62

94.78 ± 0.76

19.1

Proposed architecture

U-Net + Identity

Enhanced PatchGAN

31.47 ± 0.52

96.85 ± 0.67

14.3

Back to article page