The Deepfake Detector is an analysis tool that utilizes AI vision models to determine the authenticity of digital content.
It identifies AI-generated content through real-time overlay analysis and gallery image analysis features.
Key Features
Real-time Overlay Analysis
• Background Monitoring: Captures and analyzes screen content in real-time, even while other apps are running.
• Simultaneous Multi-Model Inference: Supports both face analysis and full-screen analysis models simultaneously.
• Dynamic Guide Provision: Updates the overlay panel in real-time with 4-level confidence indicators and detailed status descriptions based on analysis results.
Gallery Image Analysis
• Multi-Object Detection: Individually detects multiple faces contained within an image and independently calculates the tampering probability for each object.
• Visualization Overlay: Renders and displays color-coded boxes based on the analysis score in detected areas.
Technical Optimization and Security
• On-device Processing: All data analysis is completed within the device, and no information is transmitted to or collected from external servers.
• Image Filtering Control: Allows adjustment of the analysis environment, such as sharpness filtering, lighting correction, and noise reduction, to improve analysis accuracy.