Synthesize-It-Classifier: Learning a Generative Classifier through RecurrentSelf-analysis

TL;DR

This paper introduces Synthesize-It-Classifier (STIC), a novel method that enables image classifiers to generate high-quality, diverse images without an explicit generator, by iteratively refining class boundaries through synthesized samples.

Contribution

The paper presents a new generative approach for classifiers that synthesizes images via boundary-guided gradient ascent, improving both classification and image quality without traditional generative models.

Findings

STIC produces high-resolution, photo-realistic images at scale.

Iterative training improves classification accuracy and image synthesis quality.

Score-STIC enhances results on datasets like ImageNet, LSUN, and CIFAR 10.

Abstract

In this work, we show the generative capability of an image classifier network by synthesizing high-resolution, photo-realistic, and diverse images at scale. The overall methodology, called Synthesize-It-Classifier (STIC), does not require an explicit generator network to estimate the density of the data distribution and sample images from that, but instead uses the classifier's knowledge of the boundary to perform gradient ascent w.r.t. class logits and then synthesizes images using Gram Matrix Metropolis Adjusted Langevin Algorithm (GRMALA) by drawing on a blank canvas. During training, the classifier iteratively uses these synthesized images as fake samples and re-estimates the class boundary in a recurrent fashion to improve both the classification accuracy and quality of synthetic images. The STIC shows the mixing of the hard fake samples (i.e. those synthesized by the one hot class conditioning), and the soft fake samples (which are synthesized as a convex combination of classes, i.e. a mixup of classes) improves class interpolation. We demonstrate an Attentive-STIC network that shows an iterative drawing of synthesized images on the ImageNet dataset that has thousands of classes. In addition, we introduce the synthesis using a class conditional score classifier (Score-STIC) instead of a normal image classifier and show improved results on several real-world datasets, i.e. ImageNet, LSUN, and CIFAR 10.