Facial Pose Feature Flipping for Robust Face Image Rotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems face challenges in accuracy due to external environmental factors such as illumination and image composition, particularly facial pose, despite the application of deep learning.
Innovation Solution
A facial image generating method and apparatus that utilizes a convolutional neural network (CNN) to determine an assistant feature vector by applying flipped first feature vectors, position, and rotation information, and performs symmetric transformations to improve facial recognition accuracy by generating output images with varied facial poses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning is applied to facial recognition, then accuracy is improved, but recognition rate remains low due to external environmental factors and image composition
Solution Approach 1:
The system performs preliminary actions by generating multiple synthetic facial images with different poses, illuminations, and expressions before the actual recognition process. This pre-processing step creates a more comprehensive dataset that accounts for various environmental conditions, thereby improving recognition reliability without sacrificing accuracy
Solution Approach 2:
The system creates copies of the input facial image through generative models, producing multiple variations with different poses and lighting conditions. These synthetic copies are then used to train or augment the recognition system, allowing it to recognize faces under diverse conditions while maintaining high accuracy
2Adaptability or versatility
If facial recognition is performed on images with various poses and illuminations, then adaptability is improved, but measurement precision deteriorates
Solution Approach 1:
The system generates synthetic training images with various poses and illuminations in advance, creating a comprehensive dataset that covers diverse conditions. This preliminary preparation allows the recognition model to learn robust features that maintain precision across different environments
Solution Approach 2:
The system varies parameters such as pose angle, illumination direction, and expression in the generated synthetic images. By training on images with these changed parameters, the system learns to maintain recognition precision across different conditions, achieving both adaptability and accuracy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A processor-implemented facial image generating method includes: determining a first feature vector associated with a pose and a second feature vector associated with an identity by encoding an input image including a face; determining a flipped first feature vector by flipping the first feature vector with respect to an axis in a corresponding space; determining an assistant feature vector based on the flipped first feature vector and rotation information corresponding to the input image; determining a final feature vector based on the first feature vector and the assistant feature vector; and generating an output image including a rotated face by decoding the final feature vector and the second feature vector based on the rotation information.