Adaptive Hypersphere Radius for Face Recognition Training Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing face recognition systems face low overall training efficiency due to poor classification effects at the beginning stage of the classification model training process, especially when adjusting model parameters, leading to difficulties in achieving accurate feature mapping and classification.
Innovation Solution
The proposed method involves a classification model training process using a neural network that includes projecting input data into a hypersphere feature projection space with an adaptively learned hypersphere radius and dynamically adjusting the margin value based on classification effects, combined with back propagation optimization to continuously decrease the model's loss, thereby improving training efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional classification model training is used with fixed hyperparameter settings, then the training process is simple, but the classification effect is poor at the beginning stage resulting in low overall training efficiency
Solution Approach 1:
The patent applies dynamics by making the hyperparameters (hyperphere radius and margin value) adjustable and adaptive during the training process. The system dynamically changes these parameters based on training progress and performance metrics, transitioning from fixed to variable settings to optimize both early-stage classification效果和overall training efficiency
Solution Approach 2:
The patent implements parameter changes by systematically adjusting hyperparameters including the hyperphere radius and margin value during training. These parameter modifications are made in response to training performance, allowing the model to adapt to different training stages and improve classification effectiveness while maintaining efficient training
2Measurement precision
If the hyperparameters are adjusted frequently to improve classification effect, then the classification accuracy improves, but the training complexity increases
Solution Approach 1:
The patent employs feedback mechanisms where the system monitors training performance metrics and uses this information to guide hyperparameter adjustments. The loss function and classification performance provide continuous feedback that triggers parameter changes only when necessary, balancing accuracy improvement with training simplicity
Solution Approach 2:
The training system performs self-adjustment of hyperparameters based on its own performance metrics. The model automatically modifies its own hyperparameters (hyperphere radius, margin value) during training without requiring complex external intervention, reducing training complexity while maintaining accuracy
Data Source
AI summary
This application relates to a face recognition method performed at a computer server. After obtaining a to-be-recognized face image, the server inputs the to-be-recognized face image into a classification model. The server then obtains a recognition result of the to-be-recognized face image through the classification model. The classification model is obtained by inputting a training sample marked with class information into the classification model, outputting an output result of the training sample, calculating a loss of the classification model in a training process according to the output result, the class information and model parameters of the classification model, and performing back propagation optimization on the classification model according to the loss.