Total Generate: Cycle in Cycle Generative Adversarial Networks for Generating Human Faces, Hands, Bodies, and Natural Scenes

TL;DR

This paper introduces C2GAN, a novel cycle-based GAN framework that jointly exploits image and guidance data to generate realistic human faces, hands, bodies, and scenes, improving over previous models.

Contribution

The paper presents a unified cycle-in-cycle GAN architecture with two interconnected generators and three cycles, enabling enhanced cross-modal image generation and reconstruction.

Findings

Outperforms state-of-the-art models in generating realistic images

Effective in four guided image-to-image translation tasks

Provides a robust training framework with implicit cycle constraints

Abstract

We propose a novel and unified Cycle in Cycle Generative Adversarial Network (C2GAN) for generating human faces, hands, bodies, and natural scenes. Our proposed C2GAN is a cross-modal model exploring the joint exploitation of the input image data and guidance data in an interactive manner. C2GAN contains two different generators, i.e., an image-generation generator and a guidance-generation generator. Both generators are mutually connected and trained in an end-to-end fashion and explicitly form three cycled subnets, i.e., one image generation cycle and two guidance generation cycles. Each cycle aims at reconstructing the input domain and simultaneously produces a useful output involved in the generation of another cycle. In this way, the cycles constrain each other implicitly providing complementary information from both image and guidance modalities and bringing an extra supervision gradient across the cycles, facilitating a more robust optimization of the whole model. Extensive results on four guided image-to-image translation subtasks demonstrate that the proposed C2GAN is effective in generating more realistic images compared with state-of-the-art models. The code is available at https://github.com/Ha0Tang/C2GAN.