Facial Feature Synthesis Using Deep Convolutional Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current facial feature adding methods are inefficient, either requiring time-consuming three-dimensional modeling from multiple two-dimensional images or producing images that significantly differ from real pictures, and lack effective integration of deep learning techniques for improved accuracy.
Innovation Solution
A method and apparatus that generate a synthesized facial image by superimposing a feature image onto a given facial image using a deep convolutional network, with face satisfaction scores and L1 norm calculations to update network parameters, enabling efficient and accurate feature addition without relying on three-dimensional models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If three-dimensional modeling is used to add facial features, then the added features have good spatial consistency, but the processing time is long and efficiency is low
Solution Approach 1:
The patent uses two-dimensional map annotations to copy facial feature positions and directly applies them to the target image, avoiding the complex three-dimensional modeling process while maintaining reasonable spatial accuracy for the intended application
Solution Approach 2:
The patent extracts only the essential feature addition function from the complete three-dimensional modeling pipeline, using simplified two-dimensional coordinate mapping to achieve the core goal without unnecessary computational overhead
2Productivity
If two-dimensional map method is used to add facial features, then the processing is simple and fast, but the resulting image has significant differences from the real picture
Solution Approach 1:
The patent introduces a feedback mechanism where the synthesized image is processed through a deep convolutional network for face determination, and the network parameters are updated based on the satisfaction score to improve image fidelity iteratively
Solution Approach 2:
The patent replaces the manual two-dimensional map annotation method with an automated deep learning system that uses convolutional neural networks to synthesize and evaluate facial features, significantly improving image realism
3Manufacturing precision
If deep convolutional network is used for face determination and parameter updates, then the image fidelity improves, but the computational complexity increases
Solution Approach 1:
The patent segments the computational task into distinct modules: a deep convolutional network for face determination, a separate synthesis network for generating facial features, and an optimization module that coordinates their interaction through loss functions and parameter updates
Data Source
AI summary
Provided is a facial feature adding method, a facial feature adding apparatus, and a facial feature adding device. The facial feature adding method comprises: generating an image to be superimposed based on a given facial image and a feature to be added on the given facial image; and superimposing the image to be superimposed and the given facial image to generate a synthesized facial image. In addition, the facial feature adding method further comprises: generating a first face satisfaction score and a second face satisfaction score by use of a deep convolutional network for face determination and based on the synthesized facial image and a real image with the feature to be added; calculating an L1 norm of the image to be superimposed; and updating parameters of networks based on the first face satisfaction score, the second face satisfaction score, and the L1 norm.