AI Art Test Object Meta-Information Extraction for Model Training

Solution Overview

Problem

The challenge of training artificial intelligence models for art psychological analysis is exacerbated by the lack of sufficient and high-quality training data, leading to inefficiencies and inaccuracies due to the diverse and subjective nature of art expressions, necessitating a method to limit the dimension and range of training data.

Innovation Solution

A meta information extraction device and system that utilizes a user interface, first and second feature extraction units, and a library to extract and process objects and meta information from user-generated picture data, enabling efficient training of an AI model for art psychological analysis, even for users lacking drawing skills.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If diverse and subjective picture data is used for training AI model, then the model can handle various art expressions, but it takes a lot of time and cost to collect training data and the learning accuracy decreases

Engineering Contradiction:

Improveability to handle various art expressionsVSAvoidtime and cost to collect training data

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent extracts specific meta-information (objects, colors, positions, sizes) from diverse picture data, transforming unstructured artistic expressions into structured training features. This extraction process enables the AI model to learn from varied art expressions efficiently without requiring extensive manual data collection and labeling.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms picture data from its original diverse and subjective form into standardized parameters (object types, color codes, spatial coordinates, dimensions). This parameterization converts qualitative artistic expressions into quantitative features that the AI model can process efficiently, resolving the contradiction between handling diversity and maintaining learning efficiency.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If diverse and subjective picture data is used for training AI model, then the model can handle various art expressions, but the learning accuracy decreases

Engineering Contradiction:

Improveability to handle various art expressionsVSAvoidlearning accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent extracts specific meta-information (objects, colors, positions, sizes) from diverse picture data, transforming unstructured artistic expressions into structured training features. This extraction process enables the AI model to learn from varied art expressions efficiently without requiring extensive manual data collection and labeling.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms picture data from its original diverse and subjective form into standardized parameters (object types, color codes, spatial coordinates, dimensions). This parameterization converts qualitative artistic expressions into quantitative features that the AI model can process efficiently, resolving the contradiction between handling diversity and maintaining learning efficiency.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If multiple-choice questions with limited dimensions are used for training, then the AI model can be trained efficiently, but the range and dimension of training data is limited

Engineering Contradiction:

Improvetraining efficiencyVSAvoidrange and dimension of training data

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal meta-information extraction framework that works across diverse picture data types. By extracting standardized features (objects, colors, positions, sizes) that can be applied to any picture data, the system achieves both training efficiency through structured data and broad adaptability to handle various art expressions and psychological test scenarios.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent transforms picture data from its original diverse and subjective form into standardized parameters (object types, color codes, spatial coordinates, dimensions). This parameterization converts qualitative artistic expressions into quantitative features that the AI model can process efficiently, resolving the contradiction between handling diversity and maintaining learning efficiency.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12424324B2Meta information extraction device of object for artificial intelligence art psychological test and art psychological analysis system and method using the same

Publication Date: 2025.09.23 ICECREAM ART CO LTD

AI summary

Provided is a meta information extraction device of an object for artificial intelligence art psychological test and an art psychological analysis system and method using the same. The meta information extraction device extracts an object and meta information of the object from picture data by means of training of an artificial intelligence model, and analyzes picture data of a user who is a test taker with a combination of meta information to analyze a psychological state of the user. A meta information extraction device of an object for an artificial intelligence art psychological test according to the present disclosure may include, a user interface for providing an interactive environment to a user who is a test taker, a first feature extraction unit which is trained by an artificial intelligence model to extract the object from picture data generated on a screen using an object selected by the user through the user interface, a second feature extraction unit which is trained by an artificial intelligence model to extract meta information of an object extracted by the first feature extraction unit, and a library which stores the object extracted by the first feature extraction unit and the meta information extracted by the second feature extraction unit.