DINOv3 Beats Specialized Detectors: A Simple Foundation Model Baseline for Image Forensics

2026-04-17

AI Summary

This work addresses the limited generalization and architectural complexity of existing image manipulation localization methods by introducing a simple yet effective forensic baseline. The proposed approach freezes the DINOv2 ViT-L backbone and integrates low-rank adaptation (LoRA) with a lightweight convolutional decoder—the first to leverage DINOv2 and LoRA for image forensics. Requiring only a minimal number of trainable parameters, the method achieves a 17.0-point average pixel-level F1 improvement under the CAT-Net protocol and attains an F1 score of 0.774 under the MVSS-Net protocol, significantly outperforming specialized detectors. Moreover, it demonstrates strong robustness against common perturbations such as noise, JPEG compression, and blurring, confirming that pretrained visual representations encode rich forensic signals.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Cyber-crime defenses and forensics
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

With the rapid advancement of deep generative models, realistic fake images have become increasingly accessible, yet existing localization methods rely on complex designs and still struggle to generalize across manipulation types and imaging conditions. We present a simple but strong baseline based on DINOv3 with LoRA adaptation and a lightweight convolutional decoder. Under the CAT-Net protocol, our best model improves average pixel-level F1 by 17.0 points over the previous state of the art on four standard benchmarks using only 9.1\,M trainable parameters on top of a frozen ViT-L backbone, and even our smallest variant surpasses all prior specialized methods. LoRA consistently outperforms full fine-tuning across all backbone scales. Under the data-scarce MVSS-Net protocol, LoRA reaches an average F1 of 0.774 versus 0.530 for the strongest prior method, while full fine-tuning becomes highly unstable, suggesting that pre-trained representations encode forensic information that is better preserved than overwritten. The baseline also exhibits strong robustness to Gaussian noise, JPEG re-compression, and Gaussian blur. We hope this work can serve as a reliable baseline for the research community and a practical starting point for future image-forensic applications. Code is available at https://github.com/Irennnne/DINOv3-IML.

Problem

• image forensics
• fake image localization
• generalization
• manipulation detection
• deep generative models

Innovation

• DINOv3
• LoRA
• image forensics
• foundation model
• tampering localization