iFlowGAN: An Invertible Flow-Based Generative Adversarial Network for Unsupervised Image-to-Image Translation
IEEE Transactions on Pattern Analysis and Machine Intelligence(2022)
摘要
We propose iFlowGAN that learns an
invertible flow
(a sequence of invertible mappings) via
adversarial learning
and exploit it to transform a source distribution into a target distribution for
unsupervised image-to-image translation
. Existing GAN-based generative model such as CycleGAN [1], StarGAN [2], AGGAN [3] and CyCADA [4] needs to learn a highly under-constraint forward mapping
$\mathcal {F}: X \rightarrow Y$
from a source domain
$X$
to a target domain
$Y$
. Researchers do this by assuming there is a backward mapping
$\mathcal {B}: Y \rightarrow X$
such that
$\boldsymbol{x}$
and
$\boldsymbol{y}$
are fixed points of the composite functions
$\mathcal {B} \circ \mathcal {F}$
and
$\mathcal {F} \circ \mathcal {B}$
. Inspired by zero-order reverse filtering [5], we (1) understand
$\mathcal {F}$
via contraction mappings on a metric space; (2) provide a simple yet effective algorithm to present
$\mathcal {B}$
via the parameters of
$\mathcal {F}$
in light of Banach fixed point theorem; (3) provide a Lipschitz-regularized network which indicates a general approach to compose the inverse for arbitrary Lipschitz-regularized networks via Banach fixed point theorem. This network is useful for image-to-image translation tasks because it could save the memory for the weights of
$\mathcal {B}$
. Although memory can also be saved by directly coupling the weights of the forward and backward mappings, the performance of the image-to-image translation network degrades significantly. This explains why current GAN-based generative models including CycleGAN must take different parameters to compose the forward and backward mappings instead of employing the same weights to build both mappings. Taking advantage of the Lipschitz-regularized network, we not only build iFlowGAN to solve the redundancy shortcoming of CycleGAN but also assemble the corresponding iFlowGAN versions of StarGAN, AGGAN and CyCADA without breaking their network architectures. Extensive experiments show that the iFlowGAN version could produce comparable results of the original implementation while saving half parameters.
更多查看译文
关键词
Flow,bijection,unsupervised image-to-image translation,banach fixed point theorem
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络