Multi-Model Image Detection for Fake Part Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image detection methods based on deep learning are not interpretable and fail to accurately identify fake images or videos due to the loss of detailed features during image resizing and the inability to align images effectively across individuals, limiting their effectiveness in digital evidence identification.
Innovation Solution
The method employs multiple feature representation determination models trained for different parts of an object to determine feature representations of a target object in an image, allowing for comprehensive analysis and improved accuracy by utilizing more image information without resizing, and includes an authenticity evaluation model to determine the authenticity of the image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional deep learning image detection methods are used, then automation is achieved, but interpretability deteriorates and detection accuracy deteriorates due to loss of detailed features
Solution Approach 1:
The patent divides the image detection task into multiple parts by training separate feature representation determination models for different parts of the reference object. Each model processes specific regional features, preserving detailed information while maintaining automated detection capability. This segmentation allows the system to retain fine-grained feature data that would otherwise be lost in conventional holistic approaches.
Solution Approach 2:
The patent introduces a multi-dimensional feature representation space by determining feature representations at different levels of granularity. Instead of a single flattened feature vector, the system maintains hierarchical feature representations that preserve both global context and local detailed features, thereby preventing information loss while automating detection.
2Productivity
If image resizing is applied for detection, then processing efficiency is improved, but measurement precision deteriorates due to loss of detailed features
Solution Approach 1:
The patent segments the image processing into multiple resolution levels, maintaining both resized versions for efficiency and original detailed versions for precision. Different feature representation determination models process features at different scales, allowing the system to achieve both processing efficiency through resized images and measurement precision through detailed feature analysis.
Solution Approach 2:
The patent changes the processing parameters by using multiple feature representation determination models that operate at different feature extraction stages. This allows the system to process images at various resolution levels simultaneously, balancing processing efficiency with detection accuracy by selecting appropriate models for different tasks.
3Device complexity
If conventional detection methods are used, then device complexity is reduced, but reliability deteriorates due to inability to align images effectively across individuals
Solution Approach 1:
The patent segments the detection system into multiple specialized feature representation determination models, each trained for specific parts of the reference object. This segmentation improves reliability by allowing precise alignment and comparison of specific features across individuals, while the modular architecture keeps device complexity manageable through clear functional separation.
Solution Approach 2:
The patent applies local quality by training different feature representation determination models for different parts of the reference object. Each model specialized in specific regions (e.g., face, body, limbs) can align and compare local features effectively across individuals, improving overall detection reliability without requiring the entire system to be overly complex.
4Measurement precision
If multiple feature representation determination models are used, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the detection function into multiple independent feature representation determination models, each handling specific parts of the reference object. This segmentation improves measurement precision by allowing specialized analysis of different features while managing device complexity through modular design, where each model can be trained and optimized independently.
Solution Approach 2:
The patent creates a universal detection framework where multiple feature representation determination models work together within a unified system. The common framework and shared infrastructure reduce the actual complexity increase, while the multi-functional capability of analyzing different object parts improves measurement precision comprehensively.
Data Source
AI summary
Example embodiments of the present disclosure provide an image detection method, an electronic device and a computer-readable storage medium. The image detection method includes the following. Am image to be detected including a target object is obtained and multiple feature representation determination modules are obtained. The multiple feature representation determination modules are trained for different parts of a reference object, using a reference image including the reference object and an authenticity of the reference image. Multiple feature representations for different parts of the target object are determined based on the image to be detected and the multiple feature representation determination modules. An authenticity of the image to be detected is determined based on the multiple feature representations.