This week’s EYE ON NPI (video) is super-brainy but also good looking (or maybe it’s good at looking) – the Intel RealSense Camera Depth module is ready for embedding in your product to add instant facial recognition with excellent quality hardware and software from Intel.

Long-time viewers of EYE ON NPI know that we’re suckers for sensors and edge computing AI products – devices with machine learning capabilities that can function without the use of Cloud Computing or internet connectivity. Why?

1) More security controls: if it can’t connect to the internet, you don’t have to worry about data being used elsewhere by people who aren’t supposed to have it.

2) More privacy controls: if it can’t connect to the internet, you don’t have to worry about data being used elsewhere by the people who are supposed to have it but maybe are planning on selling it or misusing it.

3) Installation and usage flexibility: can install it in places without good network connectivity, don’t have to worry about a Verizon outage taking your product down.

There’s also other important parameters that may be improved by edge computing: lower power, smaller PCB, lower cost, no subscription fees. All good stuff!

This platform is available in two packages: one is a slick USB-connected rectangle that can be mounted with a ‘camera tripod’ 1/4-20 screw, and one that’s a bare PCB with a FPC connector.

That’s really neat in our opinion because it means you can use this technology with an every-day computer by plugging it into a USB port or doing a full integration into your product design.

We picked up one of each. Of course, if you’re just getting started, you may want to go with the all-enclosed one since its so easy to get going. We downloaded the Windows enrolling software (you can also enroll via the API, but this is a nice app) and updated the firmware. Then we practiced enrolling. Enrolling a face is always going to be a little annoying because it really needs to get good data.

You really want to be close up, and have your head straight-on to the camera. It took a few tries, however once done it worked great during detection even when far away.

We previous covered RealSense on EYE ON NPI, but that was with a LIDAR camera. This uses two every-day cameras to do depth calculation, so it’s a lot cheaper and smaller. The module is only a few mm thick!

The thing that really sold us on this module is the simplicity of usage – once you’ve got people enrolled, everything is done through a serial port/UART. That makes it incredibly easy to integrate into any kind of system – from a microcontroller to a single board Linux computer to a desktop running MacOS. And there’s a full API for RealSense on the Intel GitHub repo. The 54-pin FPC connector on the module exposes pretty much everything on the PCB assembly.

Best of all, the price of the full products is really low – it would be hard to DIY your own solution for less than the $87 each module costs. And with the heavy lifting of the face recognition done completely on this device, you can reduce the complexity of the rest of your design. Like, literally you could integrate this into your design in less than a week.

Both the Intel RealSense F455 starter kit and the bare bones Intel RealSense F450 PCB are available and in stock right now at Digi-Key! Order today and add solid, dependable, secure face-recognition to your product by tomorrow afternoon.

See on Digi-Key at https://www.digikey.com/short/3755t30v

Check out Intel RealSense at https://twitter.com/IntelRealSense