By Sneha Jha
Overview of Amazon's Rekognition product. (source)
Amazon’s image and video analysis platform, called Rekognition, has been available to use for a while now. Like Google’s Vision API, Microsoft’s Face API, and other services it is a suite of tools to easily perform computer vision tasks as described on their website:
“You just provide an image or video to the Rekognition API, and the service can identify the objects, people, text, scenes, and activities, as well as detect any inappropriate content. Amazon Rekognition also provides highly accurate facial analysis and facial recognition on images and video that you provide. You can detect, analyze, and compare faces for a wide variety of user verification, people counting, and public safety use cases.”
Those facial recognition features recently got it into the spotlight when the ACLU published a report saying 28 members of Congress were incorrectly matched with mugshots of people who have been arrested for a crime. The ACLU had also previously reported on its concerns about Amazon’s “dangerous new facial recognition technology” in May.
Facial recognition search with rekognition. (source)
Using Rekognition, the ACLU created a face database and search tool of 25,000 publicly available mugshots. They then searched this set of 25,000 mugshots using public photos of every current member of the House and Senate as inputs, with the the default match settings that Amazon sets for Rekognition. The Rekognition-based tool incorrectly matched 28 of the 435 members of Congress as criminals.
Overview of ACLU's examination of Rekognition. (source)
More importantly, people of color were flagged a lot more often (39%) as compared to white people (5%). It is worth noting that only 20% of the Congress members are people of color.
The biased results found by ACLU. (source)
It was not long before Amazon responded and refuted the claims.
Read On ->