Multi-model system detects deepfake images with visual explanations
MSU Team at the Explainable Deepfake Detection Challenge 2026: Grounded Artifact Evidence for Deepfake Detection
Computer Vision and Pattern Recognition
Summary
Fake images made by AI can look very real, which makes it hard to tell if a photo is real or fake. The authors created a system that not only detects fake images accurately but also highlights the parts of the image that show evidence of manipulation. Their method uses multiple AI models working together to recognize subtle clues and then explains its decisions in simple and detailed language. This helps people understand why the image was flagged as fake or real.
What this means in practice
- •For online content moderators: Automatically identify and explain fake images in social media posts to support informed moderation decisions.
- •For digital forensics teams: Use artifact evidence maps generated by AI to visually confirm and communicate image tampering in investigations.
Authors
Artem Filippov, Aleksandr Gushchin, Kirill Koltsov, Dmitriy Vatolin, Anastasia Antsiferova
Abstract
Recent advances in generative image models have made many manipulated images highly realistic, raising the need for detectors that are not only accurate but also able to provide visual evidence for their decisions. In this paper, we present our solution to the Explainable Deepfake Detection Challenge [2] on the XPlainVerse dataset [1], where systems are required to predict whether an image is real or fake and generate both complex and simple explanations grounded in visible forensic cues. Our method follows a modular detection-and-explanation design. For the real/fake decision, we build a multi-backbone detector that combines several DINOv3 models with Mesorch manipulation-localization features, bringing together pretrained visual representations, DCT-aware cues, and multi-scale forensic information. To inject explanation evidence into the detector, we use a Grounding-DINO-based pseudo-mask generation pipeline that converts local artifact descriptions from training explanations into weak patch- level supervision for an Artifact Evidence Map. We further introduce a local patch-level contrastive objective that separates artifact and authenticity evidence in the detector feature space without requiring paired images or pixel-level manipulation masks. For language output, we use class-conditional Qwen3-VL models to generate complex explanations for fake and real predictions, followed by a GRPO-optimized text simplification model. The proposed methods were trained and evaluated on the challenge subset of XPlainVerse. On the full test split, our submission achieves 0.9349 detection accuracy, 0.5571 explanation score, and a 0.7456 final challenge score.