The Euclidean Space is Evil: Hyperbolic Attribute Editing for Few-shot Image Generation

TL;DR

This paper introduces Hyperbolic Attribute Editing (HAE), a novel hyperbolic space-based method for few-shot image generation that improves diversity, quality, and controllability by leveraging hierarchical relationships among images.

Contribution

The paper presents HAE, a new hyperbolic space approach for few-shot image generation, enabling better diversity and interpretability compared to Euclidean-based methods.

Findings

HAE achieves high-quality image generation with limited data.

HAE allows controllable diversity through radius adjustment in hyperbolic space.

Experiments show HAE outperforms existing methods in quality and diversity.

Abstract

Few-shot image generation is a challenging task since it aims to generate diverse new images for an unseen category with only a few images. Existing methods suffer from the trade-off between the quality and diversity of generated images. To tackle this problem, we propose Hyperbolic Attribute Editing~(HAE), a simple yet effective method. Unlike other methods that work in Euclidean space, HAE captures the hierarchy among images using data from seen categories in hyperbolic space. Given a well-trained HAE, images of unseen categories can be generated by moving the latent code of a given image toward any meaningful directions in the Poincar\'e disk with a fixing radius. Most importantly, the hyperbolic space allows us to control the semantic diversity of the generated images by setting different radii in the disk. Extensive experiments and visualizations demonstrate that HAE is capable of not only generating images with promising quality and diversity using limited data but achieving a highly controllable and interpretable editing process.