notesum.ai

Published at November 6

Content-Style Learning from Unaligned Domains: Identifiability under Unknown Latent Dimensions

cs.LG

cs.AI

Released Date: November 6, 2024

Authors: Sagar Shrestha¹, Xiao Fu¹

Aff.: ¹School of Electrical Engineering and Computer Science, Oregon State University, Corvallis, OR 97331, USA

Arxiv: http://arxiv.org/abs/2411.03755v1

Refer to caption

Method	FID ( $\downarrow$ )			Style Diversity ( $\uparrow$ )
Method	AFHQ	CelebA-HQ	CelebA-7	AFHQ	CelebA-HQ	CelebA-7
Transitional-cGAN	38.00	8.12	70.45	–	–	–
StyleGAN-ADA	8.17	5.89	72.10	–	–	–
I-GAN	6.28	5.91	5.18	0.16 $\pm$ 0.02	0.07 $\pm$ 0.03	0.07 $\pm 0.03$
Proposed	6.19	5.70	5.27	0.50 $\pm$ 0.03	0.36 $\pm$ 0.04	0.26 $\pm$ 0.06