1.4+ billion

vehicles worldwide

1.3+ million

fatalities from car accidents annually

240 billion

estimated U.S. economic cost of accidents

clients_ThermoFisherLogo

EXECUTIVE SUMMARY

Vodafone wanted an inexpensive driver monitoring solution for off-the-shelf sale and installation. TensorIoT delivered a solution using low-cost cameras and edge computing for processing.

GOAL

How do you install driver monitoring equipment in older vehicles? And how do you manage the computational load of ML driver monitoring?

RESULTS

TensorIoT developed a cost-effective driver monitoring solution using off-the-shelf cameras and AWS Edge Machine Learning to handle the computational load.

Background

Vodafone is one of the world's largest telecommunications companies providing a wide range of services including voice, messaging and data across mobile and fixed networks. With the global connected car market expected to top $50 billion annually, and the substantial benefits both economically and socially from improved automobile safety, businesses are looking to capitalize on the em

The Challenge

Vodafone wanted to leverage their LTE platform to deploy a driver monitoring solution on their multi access edge compute (MEC) platform in order to reduce accidents and promote safer roads. However, purpose built cameras with in-built Al are expensive, hard to upgrade, and require integration at the time of vehicle purchase, making them unavailable to millions of existing car owners. Vodafone needed an inexpensive solution which could be purchased off the shelf and installed on any vehicles, old or new alike.

The Solution

TensorloT built a driver monitoring solution which works with low cost off-the-shelf cameras and offloads the computational heavy lifting to the MEC. The solution leverages Machine Learning at the Edge within AWS Greengrass to perform inference on the camera feed transmitted over the Vodafone's LTE network from the car. The solution also provides the ability to enhance safety and add new detection features over time by providing capability to seamlessly update ML modes on the edge.