AI Image Generation Using Expression Modulation Coefficients

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image generation methods using artificial intelligence models face challenges in obtaining paired images for training, especially when dealing with exaggerated expressions, and often result in erroneous outcomes.

Innovation Solution

A method and apparatus that generate images with target expressions by modulating parameters using a target expression modulation coefficient, employing a generative adversarial network and a 3D deformation statistical model to adjust expressions, enabling batch generation of paired images for training.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional AI models are used to add expression effects to images, then expression adjustment can be achieved, but the method fails when dealing with exaggerated expressions and produces erroneous outcomes

Engineering Contradiction:

Improveexpression adjustment accuracyVSAvoidcapability to handle exaggerated expressions

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent changes the parameter representation from direct expression labels to expression modulation coefficients that multiply with base expressions. This parameter transformation allows the model to handle exaggerated expressions by adjusting the magnitude of modulation coefficients rather than relying on fixed expression categories, thereby improving both accuracy and adaptability.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces a new dimension of expression control by separating expression into base expression and modulation coefficient components. This dimensional decomposition allows independent control of expression intensity and type, enabling the model to handle exaggerated expressions that conventional single-dimension approaches cannot manage.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If paired images are generated for training, then model accuracy improves, but the process is time-consuming and inefficient

Engineering Contradiction:

Improvetraining data qualityVSAvoidimage generation time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary action by pre-computing and storing expression modulation coefficients and their corresponding base expressions in a database before actual training. This allows the training process to directly query and use pre-prepared paired images rather than generating them in real-time, significantly reducing training time while maintaining data quality.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent creates copies of expression patterns by storing multiple expression modulation coefficient-image pairs in a database. During training, these pre-copied pairs can be directly retrieved and used, eliminating the need for time-consuming real-time generation and accelerating the training process without sacrificing measurement precision.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20250363699A1Image generation method and apparatus, and electronic device and storage medium

Publication Date: 2025.11.27 BEIJING ZITIAO NETWORK TECH CO LTD

AI summary

An image generation method and apparatus, an electronic device, and a storage medium are provided. A second target parameter is generated based on a first target parameter corresponding to a first image and a target expression modulation coefficient, and the second target parameter is input into an image generation model to generate a second image, so that the second image matching the first image and having a target expression can be obtained.