Table 16 Minority class performance comparison across GAN architectures. Significant values are in bold.
GAN architecture | Minority class F1 (Before) | Minority class F1 (After) | Improvement | Improvement percentage (%) |
|---|---|---|---|---|
Dual-Gland GAN (Ours) | 0.777 | 0.896 | + 0.119 | + 15.3 |
ECP-IGANN38 | 0.777 | 0.876 | + 0.099 | + 12.8 |
GSIP-GAN37 | 0.777 | 0.864 | + 0.087 | + 11.2 |
MCI-GAN40 | 0.777 | 0.851 | + 0.074 | + 9.5 |
WGAN | 0.777 | 0.842 | + 0.065 | + 8.4 |
CycleGAN | 0.777 | 0.836 | + 0.059 | + 7.6 |
Conditional GAN | 0.777 | 0.834 | + 0.057 | + 7.3 |
DCGAN | 0.777 | 0.827 | + 0.050 | + 6.4 |
Traditional GAN | 0.777 | 0.812 | + 0.035 | + 4.5 |
