AI Image Validation Using Prompt Matching and Heat Maps

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing Generative Artificial Intelligence (GAI) image generation methods require extensive user intervention and power consumption due to subjective validation processes, leading to high dissatisfaction and inefficient iteration for achieving satisfactory images.

Innovation Solution

Implement a system that includes a GAI image generator, description engine, and heat map generator, utilizing text description matching, neuroaesthetics criteria, and heat maps to automate image validation, reducing user interaction and power consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If automated validation using text description matching, neuroaesthetics criteria, and heat maps is implemented, then user intervention is reduced, but system complexity increases

Engineering Contradiction:
Improveuser interventionVSAvoidsystem complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The validation system is divided into three independent modules: text description matching module, neuroaesthetics criteria evaluation module, and heat map analysis module. Each module handles a specific aspect of image validation, allowing the complex validation task to be broken down into manageable, specialized components that can be developed and maintained independently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces multiple intermediary components including a prompt enhancement module that refines user inputs, a description engine that generates text descriptions of generated images, and a feedback mechanism that iteratively improves generation quality. These intermediaries bridge the gap between simple user input and complex validation requirements, reducing direct user intervention while managing system complexity.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If multiple validation criteria (text matching, neuroaesthetics, heat maps) are used, then image quality is improved, but processing time increases

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The system performs prompt enhancement and text description generation before the actual image generation and validation process. By preparing enhanced prompts and expected description patterns in advance, the system reduces the computational burden during iteration, allowing multiple validation criteria to be applied more efficiently without excessive processing time penalties.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements an iterative feedback loop where validation results from text matching, neuroaesthetics criteria, and heat map analysis are aggregated and used to refine both the generation process and the validation thresholds. This feedback mechanism allows the system to learn from previous validations and reduce processing time for subsequent iterations while maintaining high image quality standards.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If iterative generation with validation feedback is performed, then image accuracy is improved, but power consumption increases

Engineering Contradiction:
Improveimage accuracyVSAvoidpower consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system applies validation criteria at different levels of strictness depending on the iteration stage and image type. Not all three validation methods (text matching, neuroaesthetics, heat maps) are applied with equal intensity to every generated image. The system selectively applies validation based on confidence levels and previous iteration results, reducing unnecessary computational power consumption while maintaining image accuracy through targeted validation.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4703970A1Computer-implemented method, computer program product and computer system for image generation and validation
Publication Date: 2026.03.04 ACCENTURE GLOBAL SOLUTIONS LTD
  • EP4703970A1 patent drawingFigure 1
  • EP4703970A1 patent drawingFigure 2
  • EP4703970A1 patent drawingFigure 3

AI summary

Method, system, and computer-readable storage media for image generation and validation. Information describing features of a desired image is received and the received information is enhanced into a text prompt. The enhanced text prompt is used to generate a Generative Artificial Intelligence (GAI) image and a GAI text description of the GAI image is generated. Further, validations are performed to determine if the generated GAI image is valid or not based on a comparison of the enhanced prompt with the GAI text description, a list of predetermined neuroaesthetics criteria, and a heat map. If the generated GAI image is valid, the GAI image is used for further processing. If the generated GAI image is not valid, a process of enhancing the text prompt or generation of the GAI image is reinitiated.