Perceptual Artifacts Localization for Image Synthesis Tasks

TL;DR

This paper introduces a dataset and segmentation model for localizing perceptual artifacts in images generated by deep models, enabling automatic correction and quality assessment across various synthesis tasks.

Contribution

The study provides a new annotated dataset, a versatile segmentation model for artifact localization, and a novel inpainting pipeline for artifact rectification in image synthesis.

Findings

Effective artifact localization across multiple synthesis tasks

Model adapts to unseen models with minimal training data

Proposed pipeline improves image quality by correcting artifacts

Abstract

Recent advancements in deep generative models have facilitated the creation of photo-realistic images across various tasks. However, these generated images often exhibit perceptual artifacts in specific regions, necessitating manual correction. In this study, we present a comprehensive empirical examination of Perceptual Artifacts Localization (PAL) spanning diverse image synthesis endeavors. We introduce a novel dataset comprising 10,168 generated images, each annotated with per-pixel perceptual artifact labels across ten synthesis tasks. A segmentation model, trained on our proposed dataset, effectively localizes artifacts across a range of tasks. Additionally, we illustrate its proficiency in adapting to previously unseen models using minimal training samples. We further propose an innovative zoom-in inpainting pipeline that seamlessly rectifies perceptual artifacts in the generated images. Through our experimental analyses, we elucidate several practical downstream applications, such as automated artifact rectification, non-referential image quality evaluation, and abnormal region detection in images. The dataset and code are released.