Facial Recognition Model Using Standard Image Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Face recognition technologies face challenges in accurately recognizing faces with varying states, such as wearing hats, glasses, sunglasses, masks, or makeup, and different recognition angles due to differences in height and body shape, leading to reduced accuracy.
Innovation Solution
A method and device for facial image recognition that uses a standard face generation model to generate standard facial images from original images, trains a recognition model, and calculates a final similarity based on similarities between original and standard images to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional face recognition methods are used, then the system is simple to implement, but the recognition accuracy decreases when faces are in different states (wearing hats, glasses, masks, makeup) or at different angles
Solution Approach 1:
The system performs preliminary actions by generating standardized facial images through GANs before the actual recognition process. Multiple standard facial images representing different states (with hats, glasses, masks, makeup) and angles are pre-generated and stored. When recognition is needed, the system compares the input image against these pre-generated standards, significantly improving accuracy without adding complex real-time processing.
Solution Approach 2:
The system creates copies of facial images in standardized formats using GANs. Instead of directly recognizing the original image with all its variations and noise, the system generates idealized copies (standard facial images) that represent what the face should look like in different states. These copies serve as reference templates for accurate recognition, separating the essential facial features from state-specific variations.
2Reliability
If multiple factors (original image similarity and standard image similarity) are considered, then the recognition robustness improves, but the calculation complexity increases
Solution Approach 1:
The recognition process is segmented into two independent similarity calculation tasks: (1) calculating similarity between the input image and original facial images, and (2) calculating similarity between the input image and standard facial images. Each task can be processed separately using different weight coefficients, allowing the system to flexibly adjust the contribution of each similarity metric based on the specific recognition scenario, thereby improving robustness while managing calculation complexity.
Solution Approach 2:
The system dynamically adjusts the weight coefficients (α and β) in the final similarity calculation based on different recognition scenarios. When state variations are significant, more weight is given to standard facial image similarity; when original images are reliable, more weight is given to original image similarity. This parameter adjustment allows the system to adapt to different conditions without requiring completely different algorithms for each scenario.
Data Source
AI summary
A method for facial image recognition is provided. A plurality of original facial images are received. A plurality of standard facial images corresponding to the original facial images are generated through a standard face generation model. A recognition model is trained by using the original facial images and the standard facial images. The recognition model is tested by using the original facial image test set and a standard facial image test set until the recognition model recognizes that the first accuracy rate of the original facial image test set is higher than a first threshold value and the second accuracy rate of the standard facial image test set is higher than a second threshold value. The original facial image test set is composed of the original facial images obtained by sampling, and the standard facial image test set is composed of the standard facial images obtained by sampling.