Algorithm Comparison and Evaluation of GAN Models Based on Image Transferring from Desert to Green Field

Table 2

Values of FID, LPIPS, and runtime of experimental results of different GAN models.

Method	FID	LPIPS	Runtime (each epoch) (s)
CycleGAN	135.3	0.7051	18
GcGAN	132.2	0.7078	20
CUT	153.2	0.7056	18
FastCUT	179.4	0.7221	14
DCLGAN	174.0	0.7333	25
SimDCL	174.2	0.7520	23
LSeSim	244.7	0.7101	22

LPIPS, learned perceptual image patch similarity. Bold numbers represent the optimal values for the corresponding indicators.