Text-to-image synthesis method evaluation based on visual patterns

TL;DR

This paper proposes a new evaluation method for text-to-image synthesis that assesses realism, variety, and semantic accuracy by visualizing image representations and classifying them into visual concepts.

Contribution

It introduces a novel evaluation framework combining t-SNE visualization and classification to better measure semantic fidelity in generated images.

Findings

The method effectively gauges semantic accuracy of generated images.

Visualization of image representations reveals quality and diversity.

Classification accuracy correlates with human judgment of image quality.

Abstract

A commonly used evaluation metric for text-to-image synthesis is the Inception score (IS) \cite{inceptionscore}, which has been shown to be a quality metric that correlates well with human judgment. However, IS does not reveal properties of the generated images indicating the ability of a text-to-image synthesis method to correctly convey semantics of the input text descriptions. In this paper, we introduce an evaluation metric and a visual evaluation method allowing for the simultaneous estimation of the realism, variety and semantic accuracy of generated images. The proposed method uses a pre-trained Inception network \cite{inceptionnet} to produce high dimensional representations for both real and generated images. These image representations are then visualized in a $2$-dimensional feature space defined by the t-distributed Stochastic Neighbor Embedding (t-SNE) \cite{tsne}. Visual concepts are determined by clustering the real image representations, and are subsequently used to evaluate the similarity of the generated images to the real ones by classifying them to the closest visual concept. The resulting classification accuracy is shown to be a effective gauge for the semantic accuracy of text-to-image synthesis methods.