Perceptual Classifiers: Detecting Generative Images using Perceptual Features

2025-07-23

AI Summary

This paper addresses the poor generalization and weak degradation robustness in generative AI (GenAI) image detection. We propose a novel classification framework leveraging perceptual features from image quality assessment (IQA) models. Methodologically, we introduce— for the first time—the bandpass statistical feature space extracted by IQA models for GenAI image discrimination, employing a lightweight two-layer network trained with human visual perception priors. Our contributions are threefold: (1) model-agnostic design—no access to generator internals—is achieved, significantly enhancing cross-architecture generalization; (2) the method exhibits strong robustness against common degradations, including JPEG compression, Gaussian noise, and blur; and (3) it achieves state-of-the-art performance across multiple benchmark datasets, improving average detection accuracy by 3.2–7.8 percentage points over existing approaches.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Image Quality Assessment (IQA) models are employed in many practical image and video processing pipelines to reduce storage, minimize transmission costs, and improve the Quality of Experience (QoE) of millions of viewers. These models are sensitive to a diverse range of image distortions and can accurately predict image quality as judged by human viewers. Recent advancements in generative models have resulted in a significant influx of "GenAI" content on the internet. Existing methods for detecting GenAI content have progressed significantly with improved generalization performance on images from unseen generative models. Here, we leverage the capabilities of existing IQA models, which effectively capture the manifold of real images within a bandpass statistical space, to distinguish between real and AI-generated images. We investigate the generalization ability of these perceptual classifiers to the task of GenAI image detection and evaluate their robustness against various image degradations. Our results show that a two-layer network trained on the feature space of IQA models demonstrates state-of-the-art performance in detecting fake images across generative models, while maintaining significant robustness against image degradations.

Problem

• Detecting AI-generated images using perceptual features
• Assessing generalization of classifiers to unseen generative models
• Evaluating robustness against image degradations in detection

Innovation

• Uses IQA models for GenAI detection
• Trains two-layer network on IQA features
• Maintains robustness against image degradations