A Study in Dataset Distillation for Image Super-Resolution

TL;DR

This paper explores dataset distillation for image super-resolution, demonstrating that a small, distilled dataset can effectively train models with comparable quality to those trained on full datasets, thus enabling more efficient SR training.

Contribution

First systematic study applying dataset distillation to image super-resolution, analyzing pixel- and latent-space methods, and showing significant size reduction without quality loss.

Findings

Distilled dataset at 8.88% size retains near-original fidelity

Initialization and objectives significantly impact efficiency and quality

Establishes foundational insights for memory-efficient SR models

Abstract

Dataset distillation aims to compress large datasets into compact yet highly informative subsets that preserve the training behavior of the original data. While this concept has gained traction in classification, its potential for image Super-Resolution (SR) remains largely untapped. In this work, we conduct the first systematic study of dataset distillation for SR, evaluating both pixel- and latent-space formulations. We show that a distilled dataset, occupying only 8.88% of the original size, can train SR models that retain nearly the same reconstruction fidelity as those trained on full datasets. Furthermore, we analyze how initialization strategies and distillation objectives affect efficiency, convergence, and visual quality. Our findings highlight the feasibility of SR dataset distillation and establish foundational insights for memory- and compute-efficient generative restoration models.