Image Arrangement System for Illicit Content Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Content sharing systems face challenges in efficiently detecting illicit content among large sets of images due to human visual perception limitations and fatigue, leading to inaccurate identification and increased resource consumption.
Innovation Solution
A computer-implemented method that rearranges images based on computed concordance values characterizing visual detectability of illicit content features, using machine learning models and simulated annealing to optimize image ordering for improved detection accuracy and reduced user fatigue.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If users scan through large sets of images to detect illicit content, then detection coverage is improved, but user fatigue increases and detection accuracy decreases
Solution Approach 1:
The patent divides the large set of images into smaller subsets that are processed and displayed in manageable groups. This segmentation reduces the cognitive load on users by presenting images in smaller, less fatiguing batches rather than overwhelming them with entire large sets at once, thereby maintaining detection accuracy while reducing user fatigue.
Solution Approach 2:
The system performs preliminary sorting and organization of images before presentation to users. By pre-processing the image sets to create optimized viewing sequences and groupings, the system prepares the data in advance to reduce the effort required during actual scanning, thus improving ease of operation without compromising detection reliability.
2Reliability
If users scan through each image individually, then detection thoroughness is improved, but time consumption increases
Solution Approach 1:
The patent merges multiple individual image scanning tasks into unified subset evaluations. By combining adjacent images into cohesive groups that are evaluated together, the system maintains thorough detection coverage while reducing the total time required, as users can process multiple images simultaneously within a single scanning action rather than sequentially one by one.
Solution Approach 2:
The system transitions from a one-dimensional sequential scanning approach to a multi-dimensional parallel evaluation approach. By organizing images into subsets that can be processed simultaneously across different dimensions (spatial grouping, thematic clustering), the system achieves thorough detection without linearly increasing time consumption, effectively adding parallel processing capability.
3Productivity
If custom images are designed to attract viewers with curiosity-inducing content, then engagement is improved, but detection difficulty increases
Solution Approach 1:
The patent applies different processing and presentation qualities to different regions or types of images within the set. By identifying and highlighting specific local characteristics or patterns that indicate illicit content, the system maintains engagement through diverse image presentation while making detection easier by locally emphasizing suspicious features rather than treating all images uniformly.
Solution Approach 2:
The system introduces an intermediary analysis layer between the custom images and the detection process. This intermediary layer processes the images to identify patterns, anomalies, or indicators of illicit content, acting as a mediator that preserves the engaging nature of the original images while simplifying the detection task through pre-analysis and feature extraction.
Data Source
AI summary
A user request for a set of images is received via a graphical user interface, the subset of images having an initial order. A subset of neighboring images is selected, wherein the subset of neighboring images is ordered according to a first sequence. A first concordance value characterizing visual detectability of one or more illicit content features across images in the subset of neighboring images ordered according to the first sequence is computed. A second concordance value characterizing visual detectability of one or more illicit content features across images in the subset of neighboring images ordered according to a second sequence is computed. Responsive to a difference between the second concordance value and the first concordance value satisfying a predetermined condition, the subset of neighboring images is rearranged according to the second sequence. The previous steps are repeated for a plurality of additional subsets of neighboring images until each image of the set of images has been selected at least once as part of selected subsets of neighboring images to obtain a modified order of the set of images. The graphical user interface is caused to be modified to present the set of images on a grid according to the modified order.