Explainable Detection of AI-Generated Images with Artifact Localization Using Faster-Than-Lies and Vision-Language Models for Edge Devices

TL;DR

This paper introduces an explainable AI system that detects and localizes artifacts in low-resolution images using a lightweight classifier and vision-language models, achieving high accuracy and interpretability suitable for edge devices.

Contribution

It combines a novel lightweight classifier with vision-language models for artifact detection and explanation in low-resolution images, enabling real-time interpretability on edge hardware.

Findings

Achieves 96.5% accuracy on extended CiFAKE dataset.

Maintains inference time of 175ms on 8-core CPUs.

Provides interpretable artifact localization heatmaps.

Abstract

The increasing realism of AI-generated imagery poses challenges for verifying visual authenticity. We present an explainable image authenticity detection system that combines a lightweight convolutional classifier ("Faster-Than-Lies") with a Vision-Language Model (Qwen2-VL-7B) to classify, localize, and explain artifacts in 32x32 images. Our model achieves 96.5% accuracy on the extended CiFAKE dataset augmented with adversarial perturbations and maintains an inference time of 175ms on 8-core CPUs, enabling deployment on local or edge devices. Using autoencoder-based reconstruction error maps, we generate artifact localization heatmaps, which enhance interpretability for both humans and the VLM. We further categorize 70 visual artifact types into eight semantic groups and demonstrate explainable text generation for each detected anomaly. This work highlights the feasibility of combining visual and linguistic reasoning for interpretable authenticity detection in low-resolution imagery and outlines potential cross-domain applications in forensics, industrial inspection, and social media moderation.