Biometric Task Network With Expert Pooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial biometric algorithms face challenges with sparse or limited training datasets and redundant tasks across different applications, leading to inefficiencies in biometric analysis.
Innovation Solution
A multi-task network architecture comprising a common feature extraction neural network and multiple biometric analysis task-specific neural networks, enhanced by expert pooling, which shares results between tasks to improve training and inference efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple separate biometric analysis systems are used for different tasks, then each task can be performed with dedicated optimization, but the system complexity and resource requirements increase significantly
Solution Approach 1:
The patent implements a universal biometric analysis system that performs multiple tasks including face detection, attribute analysis, and recognition using a single integrated architecture. The system uses shared feature extraction components that serve all tasks, while task-specific modules handle individual requirements, achieving multi-functionality without proportionally increasing complexity
Solution Approach 2:
The system is segmented into modular components: a common feature extraction backbone that processes input images, and multiple task-specific neural network modules that receive the extracted features. This segmentation allows independent optimization of each task module while sharing the computational burden of feature extraction, resolving the contradiction between reliability and complexity
2Measurement precision
If task-specific neural networks are trained separately, then each network can be optimized for its specific task, but training time and computational resources increase
Solution Approach 1:
The patent merges the training process of multiple task-specific networks by implementing a unified loss function that combines objectives from all tasks. The system performs joint training where gradients from all tasks are aggregated and used to update shared feature extraction parameters simultaneously, reducing total training time while maintaining task-specific optimization through the unified objective function
Solution Approach 2:
The shared feature extraction backbone is designed to be universal across all biometric tasks, learning representations that are beneficial for multiple tasks simultaneously. This multi-functional design allows the system to achieve task-specific accuracy without requiring separate training processes for each task, significantly reducing training time
3Adaptability or versatility
If a comprehensive biometric analysis system processes all tasks, then all biometric requirements are met, but computational resources and processing time increase
Solution Approach 1:
The system segments the biometric analysis pipeline into a shared feature extraction stage and separate task-specific processing stages. This segmentation allows the computationally intensive feature extraction to be performed once and reused by all tasks, while only the necessary task-specific modules are activated based on requirements, reducing overall computational resource consumption while maintaining comprehensive task coverage
Solution Approach 2:
The system performs preliminary feature extraction that captures essential biometric information in a compressed representation. This preliminary processing prepares the data in advance for multiple tasks, so that when tasks are executed, they operate on pre-processed features rather than raw images, significantly reducing the computational resources needed for each individual task
Data Source
AI summary
Output can be provided from a selected biometric analysis task that is one of a plurality of biometric analysis tasks based on an image provided from an image sensor. The selected biometric analysis task can be performed in a deep neural network that includes a common feature extraction neural network, a plurality of biometric task-specific neural networks and a plurality of expert pooling neural networks that perform the plurality of biometric analysis tasks by inputting the image to the common feature extraction network to determine latent variables. The latent variables can be input to the plurality of biometric task-specific neural networks to determine a plurality of first outputs. Concatenated first output results can be formed and the concatenated plurality of first result outputs and the latent variables can be input to the plurality of expert pooling neural networks to determine one or more biometric analysis task outputs.