linty5 / PyTorch-Domain-Transfer-GAN

The re-implementation of diverse domain transfer methods

PyTorch-Domain-Transfer-GAN

The re-implementation of diverse domain transfer methods

1 Domain Transfer Conclusion

The domain transfer task is slightly different from style transfer task.

There are two main objectives of domain transfer task: multimodality and supervising ability. The multimodality means the model could map an input to diverse domains. While the supervising ability means the model needs paired data of target domain. So there are four categories for general models.

	Supervised	Unsupervised
Unimodal	pix2pix, CRN, task-specific GAN	UNIT, Coupled GAN, DTN, DiscoGAN, DualGAN, CycleGAN, StarGAN
Multimodal	pix2pixHD, vid2vid, BicycleGAN	MUNIT, FUNIT

2 About this Repo

• Pix2Pix

• CycleGAN

• StarGAN

• Other models are pendding