Style Classifiers for Authentic Content Search

Solution Overview

Problem

Current image search engines face challenges in identifying and retrieving authentic content, as they often classify images based on standard approaches that increase computational resources and latency, making it difficult to find stock images that appear naturally occurring.

Innovation Solution

A computer-based image retrieval system using style classifiers and a logistic regression model to determine the authenticity of images by processing them through a set of neural networks, which transform images into complex vector representations and map them to authenticity scores, allowing for prioritization of authentic content in search results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If standard image classification approaches are used, then images can be categorized, but computational resources increase and latency increases

Engineering Contradiction:

Improveauthenticity detection accuracyVSAvoidcomputational resources

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary action by pre-training style classifiers on large datasets of authentic and non-authentic images before actual search operations. The neural networks are pre-trained to recognize style patterns, so during runtime, the system only needs to apply these pre-trained classifiers rather than performing full training computations, significantly reducing operational computational resources while maintaining high detection accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system segments the image analysis task into multiple independent style classifiers, each specializing in detecting specific visual styles (authentic, non-authentic, illustrative, etc.). Instead of using one complex classifier for all tasks, the system divides the problem into specialized sub-classifiers that can be applied independently and whose results are combined, reducing the computational burden of each individual classifier while improving overall detection precision

Inventive Principle:

Principle #1Segmentation

2Measurement precision

If standard image classification approaches are used, then images can be categorized, but latency increases

Engineering Contradiction:

Improveauthenticity detection accuracyVSAvoidsearch latency

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary action by pre-training style classifiers on large datasets of authentic and non-authentic images before actual search operations. The neural networks are pre-trained to recognize style patterns, so during runtime, the system only needs to apply these pre-trained classifiers rather than performing full training computations, significantly reducing operational computational resources while maintaining high detection accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system segments the image analysis task into multiple independent style classifiers, each specializing in detecting specific visual styles (authentic, non-authentic, illustrative, etc.). Instead of using one complex classifier for all tasks, the system divides the problem into specialized sub-classifiers that can be applied independently and whose results are combined, reducing the computational burden of each individual classifier while improving overall detection precision

Inventive Principle:

Principle #1Segmentation

3Measurement precision

If comprehensive style classification is performed, then authenticity detection accuracy improves, but computational load increases

Engineering Contradiction:

Improveauthenticity score accuracyVSAvoidcomputational load

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system segments the image analysis task into multiple independent style classifiers, each specializing in detecting specific visual styles (authentic, non-authentic, illustrative, etc.). Instead of using one complex classifier for all tasks, the system divides the problem into specialized sub-classifiers that can be applied independently and whose results are combined, reducing the computational burden of each individual classifier while improving overall detection precision

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system uses copying by creating multiple specialized style classifier models that are trained on copied and augmented versions of training data. Each classifier copies and adapts the learned features for its specific style category, allowing the system to achieve comprehensive classification accuracy without requiring a single computationally intensive universal classifier

Inventive Principle:

Principle #26Copying

Data Source

PatentUS11017019B1Style classification for authentic content search

Publication Date: 2021.05.25 SHUTTERSTOCK

AI summary

Various aspects of the subject technology relate to systems, methods, and machine-readable media for authentic content search using style classifications. A system may be a search engine that uses a set of style classifiers to detect one or more styles associated with an image and a logistic regression model to determine a level of authenticity for the image based on the associated styles. Training images are fed to train a series of neural networks that output a set of style classifiers. An image is processed through the style classifiers to determine respective probabilities for each style classification. The results from the set of style classifiers are then input to the logistic regression model to determine an authenticity score for the image. For example, the authenticity score shows how authentic is an image (e.g., a score of 1.0 refers to 100% authenticity, whereas a score of 0.0 represents a non-authentic image).