Single Recognition Model for Occluded Face Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate 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

VSEngineering 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

Engineering Contradiction:
Improverecognition accuracyVSAvoidcomputational load
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

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.

Inventive Principle:
Principle #5Merging (Combining)

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.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Measurement precision

If multiple recognition models are used to improve recognition accuracy, then recognition accuracy is improved, but recognition speed decreases

Engineering Contradiction:
Improverecognition accuracyVSAvoidrecognition speed
Core Design Contradiction:
Measurement precisionVSSpeed

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.

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If multiple recognition models are used to handle occlusions, then recognition reliability is improved, but device complexity increases

Engineering Contradiction:
Improverecognition reliabilityVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

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.

Inventive Principle:
Principle #5Merging (Combining)

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

Engineering Contradiction:
Improveresource consumptionVSAvoidrecognition accuracy
Core Design Contradiction:
Use of energy by moving objectVSMeasurement precision

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.

Inventive Principle:
Principle #35Parameter changes

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.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3168781B1Method and apparatus for recognizing object, and method and apparatus for training recognition model
Publication Date: 2024.08.14 SAMSUNG ELECTRONICS CO LTD
  • EP3168781B1 patent drawingFigure 1
  • EP3168781B1 patent drawingFigure 2
  • EP3168781B1 patent drawingFigure 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.