Face Shape Adjustment Model Using Parameter Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video interactive applications have limited image style conversion options, failing to meet personalized user requirements for face shape adjustments.
Innovation Solution
A method and apparatus for generating a face shape adjustment image using a pre-trained model trained on minimum and maximum face parameters, allowing for adjustable face shape modifications within a defined range, enriching image editing functions and user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing image style conversion methods are used, then basic style conversion can be achieved, but the conversion types are limited and cannot meet personalized user requirements
Solution Approach 1:
The patent applies preliminary action by pre-training the face shape adjustment model with minimum and maximum face parameters before actual use. This pre-training process prepares the model in advance to handle various face shape conversion scenarios, enabling it to adapt to personalized requirements without requiring complex real-time adjustments during operation.
Solution Approach 2:
The patent utilizes parameter changes by introducing minimum face parameters and maximum face parameters as training constraints. These parameters define the range of face shape adjustments, allowing the model to generate diverse face shape conversions by varying within this parameter range, thereby enhancing adaptability without proportionally increasing system complexity.
2Adaptability or versatility
If face shape adjustment range is expanded to meet personalized requirements, then user personalization is improved, but control over adjustment degree becomes more difficult
Solution Approach 1:
The patent employs parameter changes by establishing minimum and maximum face parameters that define the adjustment range. The model learns to generate face shapes within this bounded range, ensuring that while the adjustment range is expanded for personalization, the output remains controlled and predictable through the parameter constraints.
Solution Approach 2:
The patent implements feedback through the training process where the model learns from paired examples of minimum and maximum face parameters. This feedback mechanism enables the model to understand the boundaries and relationships between different face shape parameters, improving its ability to control adjustment degrees accurately while maintaining expanded adaptability.
Data Source
AI summary
Embodiments of the present disclosure relate to a method for generating a face shape adjustment image, a model training method, an apparatus, and a device. The method for generating a face shape adjustment image includes: acquiring an original facial image; obtaining a face shape adjustment image corresponding to the original facial image by using a pre-trained face shape adjustment model, where the face shape adjustment model is obtained by training based on a first facial sample image, a second facial sample image, a minimum face parameter and a maximum face parameter, a face parameter corresponding to the first facial sample image is the minimum face parameter, and a face parameter corresponding to the second facial sample image is the maximum face parameter.