MetalGAN: Multi-Domain Label-Less Image Synthesis Using cGANs and Meta-Learning

TL;DR

MetalGAN introduces a novel cGAN and meta-learning based architecture that enables multi-domain, label-less image synthesis with minimal dataset usage, demonstrated on facial attribute transfer.

Contribution

The paper presents a new MetalGAN architecture combining cGAN and meta-learning to generate multi-domain images without hard-coded labels, using limited data.

Findings

Effective multi-domain image synthesis without labels

Validated on CelebA for facial attribute transfer

Requires only small dataset portions

Abstract

Image synthesis is currently one of the most addressed image processing topic in computer vision and deep learning fields of study. Researchers have tackled this problem focusing their efforts on its several challenging problems, e.g. image quality and size, domain and pose changing, architecture of the networks, and so on. Above all, producing images belonging to different domains by using a single architecture is a very relevant goal for image generation. In fact, a single multi-domain network would allow greater flexibility and robustness in the image synthesis task than other approaches. This paper proposes a novel architecture and a training algorithm, which are able to produce multi-domain outputs using a single network. A small portion of a dataset is intentionally used, and there are no hard-coded labels (or classes). This is achieved by combining a conditional Generative Adversarial Network (cGAN) for image generation and a Meta-Learning algorithm for domain switch, and we called our approach MetalGAN. The approach has proved to be appropriate for solving the multi-domain problem and it is validated on facial attribute transfer, using CelebA dataset.