Monday, March 02, 2020

EETimes on TrinamiX Face Recognition

EETimes article "Beam Profile Analysis Beats Facial Recognition Cheats" talks about TrinamiX facial recognition sensor:

In a recent briefing with EE Times in London, the managing director and founder of TrinamiX, Ingmar Bruder, told, “There are currently basically two ways of capturing depth: time-of-flight (ToF), and stereo, including structured light. We have figured out a third way, called beam profile analysis, for which we have several patents and own all the intellectual property. Our technology is based on back-scattering properties, creating a dense depth map using 40,000 dots for which we can measure the distance for each. Our technology is based on three components: a NIR camera module, flood illumination, and a light projector. This can sit on top of any facial recognition capability.

You are training the system based on the actual materials. In robot picking, you don’t need to pre-define the boundaries, as the system can look for the material and the boundaries. It can also sort materials, for example in recycling,” Bruder said. “One of the first products we are likely to see based on this technology is robot vacuum cleaners,” he said. “This is in evaluation at the moment, but products are likely on the market in 2021. Meanwhile, in consumer products, we will see both high-end and low-end smartphones using beam profile analysis in early 2021.


Update: Sueddeutsche Zeitung report talks about the problem of silicone masks in face recognition.

2 comments:

All comments are moderated to avoid spam and personal attacks.