Face Re-Aging Models for Identity-Preserving Images and 3D Geometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for re-aging faces in images and video frames are error-prone, tedious, and require significant manual effort, and conventional neural networks fail to preserve facial identities and often require specific image characteristics.
Innovation Solution
A computer-implemented method using a machine learning model to generate re-aging delta images and deform 3D geometry, preserving facial identities and handling varying depths, positions, and lighting conditions without requiring specific image preprocessing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual re-aging or 3D face rig techniques are used, then facial identity can be preserved, but the process becomes error-prone, tedious, and very time consuming
Solution Approach 1:
The system uses automated machine learning models that perform re-aging without requiring manual sculpting or editing. The neural network automatically processes input images to generate re-aged versions while preserving facial identity, eliminating the need for tedious manual operations and significantly improving productivity while maintaining reliability.
Solution Approach 2:
The patent replaces manual mechanical sculpting processes and traditional image editing techniques with automated neural network-based systems. The machine learning model automatically performs the re-aging transformation, substituting the manual mechanical process with an intelligent automated system that is both faster and more reliable.
2Productivity
If conventional neural networks are used for automatic re-aging, then productivity increases, but facial identities are not preserved and specific image characteristics are required
Solution Approach 1:
The patent modifies the parameters and architecture of conventional neural networks to specifically preserve facial identity. This includes using identity-preserving loss functions, incorporating facial landmark constraints, and adjusting network architecture to maintain key facial features while applying age transformations. These parameter changes enable the system to automatically process images while reliably preserving facial identities.
Solution Approach 2:
The neural network system is designed to handle diverse input images with varying characteristics, depths, and positions without requiring specific preprocessing. The model achieves universality by being trained on diverse datasets and incorporating robust feature extraction that works across different image conditions, maintaining both high productivity and facial identity preservation across various scenarios.
3Stability of the object's composition
If conventional neural networks require specific image characteristics, then processing consistency is improved, but adaptability to diverse input conditions deteriorates
Solution Approach 1:
The system dynamically adapts to different input image conditions through the machine learning model's ability to process varying depths, positions, and lighting conditions. The neural network adjusts its processing based on the specific characteristics of each input image, maintaining consistent output quality while handling diverse input varieties without requiring rigid preprocessing requirements.
Solution Approach 2:
The model performs preliminary feature extraction and normalization that prepares diverse input images for consistent processing. By pre-processing features in a standardized manner while preserving the essential characteristics needed for identity preservation, the system achieves both processing consistency and adaptability to various input conditions.
Data Source
AI summary
Techniques are disclosed for re-aging images of faces and three-dimensional (3D) geometry representing faces. In some embodiments, an image of a face, an input age, and a target age, are input into a re-aging model, which outputs a re-aging delta image that can be combined with the input image to generate a re-aged image of the face. In some embodiments, 3D geometry representing a face is re-aged using local 3D re-aging models that each include a blendshape model for finding a linear combination of sample patches from geometries of different facial identities and generating a new shape for the patch at a target age based on the linear combination. In some embodiments, 3D geometry representing a face is re-aged by performing a shape-from-shading technique using re-aged images of the face captured from different viewpoints, which can optionally be constrained to linear combinations of sample patches from local blendshape models.