Intel announces RealSense ID, an on-device facial authentication device, which will be available in the first quarter of 2021 at $99 and targeted at ATMs, smart locks, or kiosks (Kyle Wiggers/VentureBeat)
Intel today announced the latest update to RealSense, its range of depth and tracking technology created to offer machines the ability to perceive depth. It’s called RealSense ID, it’s a device-based solution that integrates active depth sensors with a machine-learning model for facial authentication.
Intel Realsense Q1 ATM Swiggers
Intel asserts that RealSense ID adapts to users when physical characteristics like glasses and facial hair change with time, and it works in different lighting conditions for those “with an array of sizes or skin tones.”
Also, read LOOK DENMARK 100K SELSOE SORENSEN
Major developments in the way AI can help to provide more efficient healthcare, which results in better outcomes for everyone. 115.2K1 Play Video
Transform 2022 Register NOW Watch on-demand NOW The latest developments in the ways AI can aid in better healthcare and better outcomes for everyone
However, numerous studies as well as VentureBeat’s own studies of benchmark data from the public have proven that the facial recognition algorithm is vulnerable to a variety of biases. One problem is that the data used to train algorithms tend to favor males and whites.
IBM discovered the 81% of those within the three collections of face images that are most frequently referenced in academic studies have skin that is lighter in color. Researchers have discovered that photographic techniques and technology are also able to favor lighter skin tones, including all kinds of sepia-tinged films to digital cameras with low contrast. This is why Amazon, IBM, Microsoft, and others have themselves imposed moratoriums on the selling of facial recognition software.
Also, read JAMF IS NOW MANAGING 20 MILLION DEVICES AROUND THE WORLD
As a response to the criticisms, Intel says RealSense ID is priced at $99 and will be available in Q1 2021. It has an anti-spoofing feature built-in to guard against fraudulent entry attempts made through videos, photos, or masks. The company claims that it has a “one-in-1-million” fake acceptance rate, and it processes facial images locally, encrypts user information, and is activated via “user consciousness.” That is it doesn’t authenticate until it is prompted by a registered user.