Table 4 Quantitative evaluation results on the CityScapes dataset (IS \(\uparrow\), FID \(\downarrow\), PI \(\downarrow\), SCL \(\downarrow\), SSS \(\uparrow\)).
From: Enhancing media image style transfer with advanced StyleGAN2 architectures
Model | IS | FID | PI | SCL | SSS |
|---|---|---|---|---|---|
CycleGAN | 7.16 | 76.83 | 0.489 | 0.423 | 0.476 |
CUT | 8.69 | 58.27 | 0.412 | 0.358 | 0.543 |
DCLGAN | 8.37 | 50.64 | 0.375 | 0.311 | 0.589 |
StyleFormer | 8.52 | 47.89 | 0.342 | 0.287 | 0.615 |
StyleDiffusion | 8.75 | 46.32 | 0.316 | 0.264 | 0.652 |
Enhanced CycleGAN | 8.63 | 45.78 | 0.328 | 0.275 | 0.637 |
Ours | 8.92 | 49.13 | 0.253 | 0.201 | 0.728 |
