Machine Learning Times
Machine Learning Times
EXCLUSIVE HIGHLIGHTS
Climate Tech Needs Machine Learning, Says PAW Climate Conference Chair
  Straight from the horse’s mouth – the founding...
Predictive Policing: Six Ethical Predicaments
  Originally published in KDNuggets. This article is based...
Measuring Invisible Treatment Effects with Uplift Analysis
  Models make predictions by identifying consistent correlations in...
Machine Learning: Business Leaders Must Take an Enlightening Look Under Its Hood (New Training Program)
  In this article, I identify unmet learner needs...
SHARE THIS:

5 months ago
Amazon Touts AI for Social Distancing Amid Worker Complaints

 
Originally published in Wired.com, June 18, 2020

Facing criticism over workplace safety, the company is using cameras, sensors, and augmented reality to warn employees when they’re too close to one another.

Covid-19 sent Amazon scrambling to adapt its logistics facilities to contain contagion while also handling a surge in orders from customers stuck at home. Some workers say it has struck the wrong balance between profits and safety.

Amazon has vigorously defended its policies and pushed back with details of some of its Covid-19 safety measures. The latest reveal is an artificial intelligence system that analyzes images from security cameras in Amazon facilities and alerts management of potential social distancing violations.

Proxemics, as the system is called, was built by AI experts in Amazon’s robotics division and deployed in mid-March. It now operates in more than 1,000 Amazon buildings around the world, one of several high-tech Amazon projects to monitor social distancing in its facilities.

The system grabs images from security cameras every few minutes and discards those without any people in view. Algorithms automatically blur out any people visible to protect privacy and select any in which people appear to be too close together, referring those to human reviewers elsewhere. Because the software is analyzing images from cameras that don’t directly measure distance, it uses the apparent size of people in the frame and the number of pixels between them to flag possible violations. When a reviewer sees cause for concern in a photo, they include the details in a regular report sent to building managers that summarizes recent social distancing violations in their facility.

To continue reading this article click here.

Leave a Reply