A Turn Toward Better Alignment: Few-Shot Generative Adaptation with Equivariant Feature Rotation

TL;DR

This paper introduces Equivariant Feature Rotation (EFR), a novel few-shot image generation method that aligns source and target domains in a self-rotated feature space, improving adaptation with limited data.

Contribution

EFR employs adaptive rotations within a Lie Group to align features, preserving structure and effectively bridging domain gaps in few-shot generative tasks.

Findings

Significantly improves generative quality on multiple datasets.

Effectively preserves intra-domain structure during adaptation.

Outperforms existing few-shot generative methods.

Abstract

Few-shot image generation aims to effectively adapt a source generative model to a target domain using very few training images. Most existing approaches introduce consistency constraints-typically through instance-level or distribution-level loss functions-to directly align the distribution patterns of source and target domains within their respective latent spaces. However, these strategies often fall short: overly strict constraints can amplify the negative effects of the domain gap, leading to distorted or uninformative content, while overly relaxed constraints may fail to leverage the source domain effectively. This limitation primarily stems from the inherent discrepancy in the underlying distribution structures of the source and target domains. The scarcity of target samples further compounds this issue by hindering accurate estimation of the target domain's distribution. To overcome these limitations, we propose Equivariant Feature Rotation (EFR), a novel adaptation strategy that aligns source and target domains at two complementary levels within a self-rotated proxy feature space. Specifically, we perform adaptive rotations within a parameterized Lie Group to transform both source and target features into an equivariant proxy space, where alignment is conducted. These learnable rotation matrices serve to bridge the domain gap by preserving intra-domain structural information without distortion, while the alignment optimization facilitates effective knowledge transfer from the source to the target domain. Comprehensive experiments on a variety of commonly used datasets demonstrate that our method significantly enhances the generative performance within the targeted domain.