Unsupervised Visual Attribute Transfer with Reconfigurable Generative Adversarial Networks

TL;DR

This paper introduces an unsupervised approach for transferring visual attributes in images without needing paired datasets, broadening the applicability of attribute transfer techniques.

Contribution

It presents a novel unsupervised generative adversarial network that learns attribute transfer without requiring corresponding image pairs, unlike previous supervised methods.

Findings

Effective attribute transfer demonstrated on various tasks

No need for paired training data

Visual results verify the method's effectiveness

Abstract

Learning to transfer visual attributes requires supervision dataset. Corresponding images with varying attribute values with the same identity are required for learning the transfer function. This largely limits their applications, because capturing them is often a difficult task. To address the issue, we propose an unsupervised method to learn to transfer visual attribute. The proposed method can learn the transfer function without any corresponding images. Inspecting visualization results from various unsupervised attribute transfer tasks, we verify the effectiveness of the proposed method.