Face Syncing in Distributed Computing via Centralized Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed computing environments, maintaining consistency of face detection results across multiple devices is challenging due to variations in face detection algorithms and hardware, leading to inconsistent results and a poor user experience.
Innovation Solution
A method and system for face syncing that involves detecting faces, generating faceprints, clustering them, and sending face crop images to a network-based distributed syncing service, allowing for robust synchronization of face detection data across devices while protecting user privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If face detection is performed on encrypted media assets stored in the online library, then user privacy is protected, but face detection results are inconsistent across different devices
Solution Approach 1:
The patent introduces an intermediary face detection service that operates on decrypted media assets in the online library. This service acts as a mediator between the encrypted storage system and the client devices, performing face detection centrally and returning consistent results to all devices. The intermediary approach maintains privacy protection through encrypted storage while ensuring reliability through centralized processing.
Solution Approach 2:
The patent merges the face detection functionality from individual client devices into a centralized online service. Instead of each device performing its own face detection on encrypted assets, the system combines face detection operations into a single centralized process that operates on decrypted assets in the online library, ensuring consistent results across all devices.
2Adaptability or versatility
If different versions of face detection algorithms are used on different devices, then device compatibility is improved, but face detection results become inconsistent
Solution Approach 1:
The patent implements a universal face detection service in the online library that serves all client devices regardless of their individual capabilities or algorithm versions. This centralized service provides a unified face detection function that returns consistent results to all devices, making the system adaptable to different device versions while maintaining detection accuracy consistency.
Solution Approach 2:
Instead of having each device adapt its face detection to match others, the patent inverts the approach by having all devices rely on a centralized service that performs detection centrally. The inversion shifts the adaptation burden from multiple devices to a single centralized service, ensuring consistency while maintaining compatibility.
3Speed
If face detection is performed locally on each device, then processing speed is improved, but synchronization of face detection results across devices becomes difficult
Solution Approach 1:
The patent performs face detection in advance on all media assets stored in the online library before they need to be accessed by client devices. This preliminary action pre-computes face detection results and stores them centrally, so when devices access the media assets, the face detection data is already available for immediate use, eliminating synchronization delays.
Solution Approach 2:
The patent creates and stores copies of face detection results in the online library that can be efficiently distributed to multiple client devices. Instead of having each device perform detection and then synchronize results, the system creates a master copy of detection results that all devices can access, dramatically reducing synchronization time.
Data Source
AI summary
Embodiments are disclosed for face syncing in a distributed computing environment. In an embodiment, a method comprises: obtaining, by a processor, media assets that include faces of individuals; detecting, by the processor of a mobile device, the faces on the media assets; generating, by the processor, faceprints for the detected faces; clustering, by the processor, the faceprints into clusters; generating, by the processor, a face crop image for each cluster; and sending, by the processor, the face crop images to a network-based, distributed syncing service.