Distributed Machine Learning for Police Recording Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Police recording devices face challenges in using machine learning models due to storage space requirements, resource intensity, and limited time for processing media files, especially in law enforcement settings where network connectivity may be unreliable.
Innovation Solution
A distributed processing system that allows recording devices to share media content and process it using machine learning models locally, eliminating the need for remote servers and enabling parallel processing among multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are used locally on recording devices, then processing speed and reliability are improved, but storage space requirements and device complexity increase
Solution Approach 1:
The system segments the machine learning processing workload across multiple recording devices. Each device can independently process portions of media files using locally stored model portions, dividing the complex processing task into manageable segments that reduce individual device complexity requirements while maintaining overall system reliability
Solution Approach 2:
Recording devices are designed with multi-functionality, serving both as media capture devices and as distributed computing nodes. The devices can store and execute portions of machine learning models, process media files locally, and communicate with other devices in the network, reducing reliance on centralized processing infrastructure
2Loss of time
If machine learning models are used locally on recording devices, then processing time is reduced, but storage space requirements increase
Solution Approach 1:
Machine learning models are divided into multiple portions that can be distributed across different recording devices. Each device stores only the portions it needs to execute, reducing individual storage requirements while enabling parallel processing that significantly reduces overall processing time
Solution Approach 2:
Portions of machine learning models are copied to multiple recording devices in the network. This allows parallel execution of processing tasks across multiple devices, reducing total processing time while each individual device maintains only a fraction of the total model storage requirement
3Device complexity
If remote servers are used for processing, then device complexity is reduced, but network dependency increases
Solution Approach 1:
Recording devices pre-load necessary portions of machine learning models into local memory before processing is needed. This preliminary action enables devices to execute processing tasks independently without requiring real-time network connectivity, ensuring reliability in remote or offline environments while maintaining manageable device complexity
Data Source
AI summary
Various embodiments of the present disclosure increase the technical utility of a recording device and local recording device system by enabling a recording device to use a machine learning model to process media content received from a separate recording device. This allows a machine learning model to be used to process media content at a local system without a remote network connection or independent of whether a remote computing system is available over a network connection. Many embodiments eliminate the need for a separate computing device altogether for purposes of using a machine learning model. Embodiments of the present disclosure also decrease the time required to complete processing of a media file by processing the media file in parallel among multiple recording devices and/or by eliminating time associated with uploading a media file to cloud-based system and receiving one or more output values back from the cloud-based system.