Augmentation-Interpolative AutoEncoders for Unsupervised Few-Shot Image Generation

Abstract

We aim to build image generation models that generalize to new domains from few examples. To this end, we first investigate the generalization properties of classic image generators, and discover that autoencoders generalize extremely well to new domains, even when trained on highly constrained data. We leverage this insight to produce a robust, unsupervised few-shot image generation algorithm, and introduce a novel training procedure based on recovering an image from data augmentations. Our Augmentation-Interpolative AutoEncoders synthesize realistic images of novel objects from only a few reference images, and outperform both prior interpolative models and supervised few-shot image generators. Our procedure is simple and lightweight, generalizes broadly, and requires no category labels or other supervision during training.

References in corpus (17)

• Generative Adversarial Networks: An Overview
• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• Data Augmentation Generative Adversarial Networks
• EMNIST: an extension of MNIST to handwritten letters
• Generalization and Equilibrium in Generative Adversarial Nets (GANs)
• Optimizing the Latent Space of Generative Networks
• Towards a Neural Statistician
• Generating Diverse High-Fidelity Images with VQ-VAE-2
• Few-Shot Unsupervised Image-to-Image Translation
• FIGR: Few-shot Image Generation with Reptile
• Generative adversarial interpolative autoencoding: adversarial training on latent space interpolations encourage convex latent distributions
• The Variational Homoencoder: Learning to learn high capacity generative models from few examples
• On Adversarial Mixup Resynthesis
• Image Generation From Small Datasets via Batch Statistics Adaptation
• Can VAEs Generate Novel Examples?