Use Classifier as Generator

TL;DR

This paper introduces a simple method to generate images directly from a trained classifier without retraining, demonstrating recognizable results on MNIST with limited quality.

Contribution

It proposes a novel approach to use a pre-trained classifier as a generator, avoiding the need for retraining or additional constraints.

Findings

Produces recognizable images on MNIST

Requires no retraining of classifier or generator

Limited image quality but effective for recognition

Abstract

Image recognition/classification is a widely studied problem, but its reverse problem, image generation, has drawn much less attention until recently. But the vast majority of current methods for image generation require training/retraining a classifier and/or a generator with certain constraints, which can be hard to achieve. In this paper, we propose a simple approach to directly use a normally trained classifier to generate images. We evaluate our method on MNIST and show that it produces recognizable results for human eyes with limited quality with experiments.