Training on Thin Air: Improve Image Classification with Generated Data

TL;DR

This paper introduces Diffusion Inversion, a method using Stable Diffusion to generate high-quality, diverse training images that significantly improve image classification performance and reduce sampling time.

Contribution

We propose Diffusion Inversion, a novel technique that leverages pre-trained generative models to create effective training data, outperforming existing methods and enhancing various neural architectures.

Findings

2-3x increase in sample efficiency

6.5x reduction in sampling time

Consistent performance improvements across datasets

Abstract

Acquiring high-quality data for training discriminative models is a crucial yet challenging aspect of building effective predictive systems. In this paper, we present Diffusion Inversion, a simple yet effective method that leverages the pre-trained generative model, Stable Diffusion, to generate diverse, high-quality training data for image classification. Our approach captures the original data distribution and ensures data coverage by inverting images to the latent space of Stable Diffusion, and generates diverse novel training images by conditioning the generative model on noisy versions of these vectors. We identify three key components that allow our generated images to successfully supplant the original dataset, leading to a 2-3x enhancement in sample complexity and a 6.5x decrease in sampling time. Moreover, our approach consistently outperforms generic prompt-based steering methods and KNN retrieval baseline across a wide range of datasets. Additionally, we demonstrate the compatibility of our approach with widely-used data augmentation techniques, as well as the reliability of the generated data in supporting various neural architectures and enhancing few-shot learning.