Synthetic Visual Media Detection via Frequency Domain Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting synthetic images and video are inefficient, costly, and require significant skill, with standard machine learning techniques lacking understanding of expected visual media formats, leading to low accuracy and reliability.
Innovation Solution
A method that identifies synthetic images and video by separating the visual media into portions containing and not containing specific features, using a trained machine learning model to analyze these portions in the frequency domain, and providing data indicating whether the media is synthetic or genuine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection by skilled individuals is used to identify synthetic visual media, then detection accuracy can be high, but the process becomes costly and time-consuming
Solution Approach 1:
The patent replaces manual inspection by skilled individuals with an automated machine learning system that uses neural networks to analyze visual media. The system automatically detects synthetic content through computational algorithms, eliminating the need for human experts to manually examine each image or video, thus reducing time and cost while maintaining detection accuracy.
Solution Approach 2:
The machine learning model is trained to autonomously detect synthetic visual media without requiring continuous human intervention or expertise. Once trained, the system independently analyzes visual media, makes detection decisions, and can be deployed at scale to process large volumes of content automatically, making the detection process self-sufficient.
2Extent of automation
If standard machine learning methods are used for detection, then automation is achieved, but accuracy and reliability remain low due to lack of understanding of expected visual media formats
Solution Approach 1:
The patent transforms the visual media from spatial domain to frequency domain using Fourier transform, changing the representation parameters of the input data. This parameter transformation allows the neural network to detect subtle artifacts and patterns in the frequency spectrum that are characteristic of synthetic media, thereby improving detection reliability while maintaining automation.
Solution Approach 2:
The system moves the analysis from the conventional spatial dimension to the frequency dimension by applying Fourier transform. This dimensional change enables the detection of patterns and artifacts that are not apparent in the spatial domain, providing the automated system with additional discriminative features to improve reliability.
3Adaptability or versatility
If existing detection methods are applied, then some level of detection is possible, but they fail to handle cases where background objects are absent or faces are not fully visible
Solution Approach 1:
The neural network detector is designed to be universal and handle multiple types of visual media including images with faces, images without faces, videos with audio, and videos without audio. The system performs the same frequency domain analysis and classification task across all these different input types, making it adaptable to various scenarios while maintaining consistent detection reliability.
Data Source
AI summary
Method and system for detecting synthetic visual media comprising the steps of receiving visual media, wherein the visual media is synthetic visual media or genuine visual media. Identifying at least one feature within the visual media. Identifying a first portion of the visual media and second portion of the visual media, the first portion of the visual media containing the at least one feature and the second portion of the visual media not including the at least one feature. Providing the first portion and the second portion to a trained machine learning (ML) model. The trained ML model providing data indicating the visual media to be synthetic and/or genuine.