Disentangling Mean Embeddings for Better Diagnostics of Image Generators

TL;DR

This paper introduces a method to analyze image generator performance by disentangling mean embeddings into pixel clusters, improving diagnostics and interpretability of model behavior across image regions.

Contribution

It proposes a novel approach using central kernel alignment to decompose similarity metrics into pixel cluster contributions, enhancing explainability of image generator evaluations.

Findings

Enables identification of specific image regions where generators underperform

Improves interpretability of generator evaluation metrics

Facilitates targeted diagnostics of model misbehavior

Abstract

The evaluation of image generators remains a challenge due to the limitations of traditional metrics in providing nuanced insights into specific image regions. This is a critical problem as not all regions of an image may be learned with similar ease. In this work, we propose a novel approach to disentangle the cosine similarity of mean embeddings into the product of cosine similarities for individual pixel clusters via central kernel alignment. Consequently, we can quantify the contribution of the cluster-wise performance to the overall image generation performance. We demonstrate how this enhances the explainability and the likelihood of identifying pixel regions of model misbehavior across various real-world use cases.