Face Alignment Model Selection for Real-Time Posture Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems face challenges in performing real-time face alignment on images with unknown postural angles, leading to high computational burden and hardware requirements, making them unsuitable for mobile devices.
Innovation Solution
A method and system for face alignment that constructs multiple models for different postural angles, iteratively evaluates their performance, and refines the alignment through a series of stages using confidence scores to determine a final shape, reducing computational load and hardware needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple face alignment models corresponding to different postural angles are constructed and operated separately, then face alignment accuracy for various postures is improved, but computational burden and hardware requirements increase significantly
Solution Approach 1:
The patent segments the face alignment task by dividing postural angles into different ranges, with each model responsible for a specific range. This allows the system to handle different postures with specialized models while avoiding the need to run all models for every input, thus reducing computational burden while maintaining alignment accuracy.
Solution Approach 2:
The patent changes the parameter of postural angle range coverage for different models. Each model is trained on a specific range of postural angles rather than all angles, which reduces the complexity of individual models and the overall computational burden while maintaining comprehensive coverage through model selection.
2Measurement precision
If multiple face alignment models are operated separately for different postural angles, then alignment accuracy is improved, but real-time processing capability deteriorates
Solution Approach 1:
The patent performs preliminary classification of the input image's postural angle range before selecting which model to operate. This preliminary action enables the system to quickly identify the appropriate model and execute only the necessary computation, achieving real-time processing while maintaining high alignment accuracy through specialized models.
Solution Approach 2:
By segmenting the model selection process based on postural angle ranges, the system avoids running multiple models simultaneously. Instead, it divides the task into stages: first classify the angle range, then select and execute the corresponding model, thereby improving processing speed while maintaining accuracy.
3Adaptability or versatility
If multiple face alignment models are used for different postural angles, then adaptability to various postures is improved, but device resource requirements increase
Solution Approach 1:
The patent segments the overall posture coverage task into multiple specialized models, each handling a specific angular range. This segmentation allows the system to achieve high adaptability across all postures while using less hardware resources per model compared to a single comprehensive model, as each model is optimized for its specific range.
Solution Approach 2:
The patent creates a universal system that can handle all postural angles through a collection of specialized models. Each model is a simplified version optimized for specific angles, and together they provide universal coverage, reducing the hardware resources needed compared to a single model attempting to handle all angles simultaneously.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method and system for face alignment. The method may include obtaining an image processing model set including M (M ≥ 2) candidate models, and obtaining a test image including a target face. The method may also include conducting T (T ≥ 1) stages of model set updating operation. Each stage of the T stages of model set updating operation may include conducting a performance evaluation to each candidate model of the image processing model set with respect to the test image, and updating the image processing model set by excluding at least one model from the image processing model set based on the performance evaluation. The method may further include designating, after completing the T stages of model set updating operation, at least one candidate model of the image processing model set as a target model, and determining, based on the target model, a result shape as a shape of the target face.