ReMix: Towards Image-to-Image Translation with Limited Data

TL;DR

ReMix is a data augmentation technique for GAN-based image-to-image translation that interpolates features and uses a perceptual content loss to improve generalization with limited data.

Contribution

It introduces feature-level interpolation and a novel perceptual relation-based content loss to enhance I2I translation under data scarcity.

Findings

Significant performance improvements on multiple tasks.

Reduces overfitting and content ambiguity.

Easy to integrate into existing GAN frameworks.

Abstract

Image-to-image (I2I) translation methods based on generative adversarial networks (GANs) typically suffer from overfitting when limited training data is available. In this work, we propose a data augmentation method (ReMix) to tackle this issue. We interpolate training samples at the feature level and propose a novel content loss based on the perceptual relations among samples. The generator learns to translate the in-between samples rather than memorizing the training set, and thereby forces the discriminator to generalize. The proposed approach effectively reduces the ambiguity of generation and renders content-preserving results. The ReMix method can be easily incorporated into existing GAN models with minor modifications. Experimental results on numerous tasks demonstrate that GAN models equipped with the ReMix method achieve significant improvements.