Single Recognition Model for Occluded Face Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems using multiple recognition models are inefficient due to increased computational load and resource consumption, leading to decreased recognition speed and accuracy, especially in the presence of occlusions such as sunglasses or masks.
Innovation Solution
A single recognition model based on a deep neural network is used to extract and recognize features from input images, including those with occlusions, by determining feature weights based on occlusion probabilities and generating occlusion maps to improve robustness and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple recognition models are used to improve recognition accuracy, then recognition accuracy is improved, but computational load and resource consumption increase
Solution Approach 1:
The patent merges multiple recognition models into a single unified recognition model. This single model integrates the functionality of multiple specialized models, reducing computational overhead and resource consumption while maintaining recognition accuracy through a unified feature extraction and classification process.
Solution Approach 2:
The single recognition model is designed to perform multiple functions that were previously handled by separate models. It can recognize faces under various conditions (different lighting, angles, occlusions) using a single versatile model, eliminating the need for multiple specialized models and reducing overall system complexity.
2Measurement precision
If multiple recognition models are used to improve recognition accuracy, then recognition accuracy is improved, but recognition speed decreases
Solution Approach 1:
By combining multiple recognition models into one, the patent eliminates the sequential processing requirement of multiple models. The single model processes each input image in one pass, significantly improving recognition speed while maintaining accuracy through integrated feature extraction and classification capabilities.
3Reliability
If multiple recognition models are used to handle occlusions, then recognition reliability is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple occlusion-handling models into a single recognition model with integrated occlusion detection and feature extraction capabilities. This unified approach maintains reliability by handling various occlusion types within one model, while reducing system complexity by eliminating the need for multiple specialized models and their coordination.
4Use of energy by moving object
If a single recognition model is used to reduce computational load, then resource consumption is reduced, but recognition accuracy may decrease
Solution Approach 1:
The patent employs parameter changes and optimization techniques within the single recognition model to maintain high accuracy. By adjusting model parameters, using efficient feature extraction methods, and optimizing the decision-making process, the single model achieves accuracy comparable to multiple models while consuming fewer resources.
Solution Approach 2:
The patent replaces the mechanical system of multiple sequential model executions with a single integrated model that uses optimized algorithms and efficient computation. This substitution maintains recognition accuracy through sophisticated single-model processing while significantly reducing computational load and resource consumption.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A method and an apparatus for recognizing an object are disclosed. The apparatus may extract a plurality of features from an input image using a single recognition model and recognize an object in the input image based on the extracted features. The single recognition model may include at least one compression layer configured to compress input information and at least one decompression layer configured to decompress the compressed information to determine the features.