Face Recognition Synchronization Between Local and Server Models

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing methods face challenges in accurately classifying images due to differences between images stored on a smartphone and those uploaded to a server, as well as limitations in face recognition precision between local and server-based models.

Innovation Solution

An image processing method that involves face recognition on both locally stored and server-stored images using different recognition models, with image matching and classification processes to synchronize and improve the accuracy of face recognition, allowing for precise classification of images based on shared face sets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If face recognition is performed using local recognition models on smartphone images, then processing speed is improved, but recognition precision deteriorates

Engineering Contradiction:

Improveprocessing speedVSAvoidrecognition precision

Core Design Contradiction:

SpeedVSMeasurement precision

Solution Approach 1:

The system divides the face recognition task into two segments: initial fast recognition using local models on the smartphone, and subsequent precise verification using server-based models. This segmentation allows the system to benefit from both fast local processing and accurate server-based recognition without requiring all images to be processed by the computationally intensive server models.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an image matching mechanism as an intermediary between local and server-based face recognition. The system first performs rapid local face recognition, then uses image matching to identify candidate images for server verification, and finally applies server-based recognition only to these candidates. This intermediary step optimizes the balance between processing speed and recognition precision.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If all images are uploaded to the server for processing, then recognition precision is improved, but loss of time increases

Engineering Contradiction:

Improverecognition precisionVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

Instead of uploading all images to the server for face recognition, the system performs partial action by first conducting local face recognition on all images, then selectively uploading only those images that require verification or contain uncertain face data to the server. This reduces the number of images processed by the server, thereby minimizing upload time and processing delays while maintaining recognition precision for critical images.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system performs preliminary face recognition using local models before determining which images need server-based processing. This preliminary action filters out images that can be confidently processed locally, so that only images requiring server verification are uploaded, thus reducing overall processing time while ensuring precision for images that need it.

Inventive Principle:

Principle #10Preliminary action

3Adaptability or versatility

If different recognition models are used for local and server images, then processing flexibility is improved, but inconsistency in classification results occurs

Engineering Contradiction:

Improveprocessing flexibilityVSAvoidclassification consistency

Core Design Contradiction:

Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The system implements a feedback mechanism where server-based face recognition results are used to verify and correct local recognition outcomes. The server compares its recognition results with local results, and when discrepancies are found, the system uses the server's more accurate results to update or correct the local classification. This feedback loop ensures consistency between local and server-based classifications while maintaining the flexibility of using different models.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent merges the results from local and server-based face recognition systems through an image matching and result comparison process. By combining the strengths of both systems—local speed and server accuracy—and reconciling their results through systematic comparison and verification, the system achieves both processing flexibility and classification consistency.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentEP3493113B1Image processing method, computer device, and computer readable storage medium

Publication Date: 2023.12.27 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD

Figure 1~3
Figure 4
Figure 5~6

AI summary

Embodiments of the disclosure provide an image processing method. The method includes: acquiring (302) a first face set by performing face recognition on a first image set according to a first recognition model; in response to detecting a connection to a network, receiving (304, 406) a second face set sent by a server, the second face set being acquired by the server through face recognition on images in a second image set according to a second recognition model, a recognition precision of the first recognition module being lower than a recognition precision of the second recognition model; performing (306) image matching on the images in the first image set and the images in the second image set; processing the first face set according to the second face set and a result of the image matching; and classifying (308, 416) the first image set according to the processed first face set.