 Did you key and a fruit present? This week's Ion MPI is an XP product Lady Aida, what is the new product? Okay from did you key this week? It's definitely giving an eye on the eye and MPI really is the new bag of eyes It's a bag of eyes. Yay Is the XP SLN? Visnas IOT I'm saying that currently it's a edge machine learning face recognition hardware that once again a completely on the edge it was on a mic controller it uses the NXP RT 10 6x series, which if you're like hey, that sounds familiar It's because it's also used in the teen C4. It's a very similar chip. It's the same hardware But this one is licensed to use I guess this Machine learning build that does machine learning for face recognition. It does a nice fast job of it. It's very effective and precise and It does not require a full Linux bill It doesn't require the internet and and because it's a microcontroller it can go into deep sleep mode So it's perfect when you want to do face recognition But you don't want to depend on the internet or if you need to be able to go to low power mode because it may Be ones on a battery or maybe it just shouldn't it doesn't it shouldn't have to depend on power so for example if you're doing a door entry system and You know, you don't want people to get locked out because there's no power if the power goes out They should still be able to get in and out This could run on battery backup power something that a Linux Computer would not be able to do nearly as easily. So let's take a look at the dev board so the chip itself is on the back I'll show that in a second, but this is the dev board you get and It's got USB-C. It's got a couple buttons in the center PIR sensor. That's the white thing and then two Cameras on the right and let's look at this in more detail. So There's the front the back and they kind of plug together. So on the left hand side is the microcontroller. So the microcontroller is again that RT 10x series It's got hyper flash memory memory for this chip is stored on flash I don't think it has internal flash memory or if it does it needs more to run this code SD RAMs you have plenty of memory. You're gonna need a lot of memory to do Any kind of visual recognition because you have to get so much visual data and analyze it There's an SWD port. There's mono output the example I don't think uses it But if you want to reprogram the code and add audio output, of course, you get that there's also PDM Microphones so you can also add in some voice recognition if you need a PIR sensor The dev board has four buttons which are used in the example code There's a boot mode. So if you want to reprogram it with the built-in bootloader lithium-ion batteries so you can power it off of battery power and not completely indicated here, but you see those FPC connectors. There is an external TFT display that you can plug in sold separately and there's a USB C connector on the bottom Great. So if you take that USB C and you plug it in This shows up as a comp or and a camera. It was me like it actually shows up as a device Because it's a microcontroller and it can do stuff like that You don't have to deal with like gadget mode with Linux. It just shows up as like literally a camera So you open up your camera software like in this case I'm running like a Microsoft camera or something and it comes up and it has something like this It has the visual output of what it's seeing There's like a sort of a blue square in the center telling you this is what it's looking at for the face recognition area I turn on the debug so that text overlay that tells you like what the output of the various processing stages are On the bottom it tells you what the application is and this is the electronic lock application where you don't recognize as a face It unlocks something There's at the bottom you can barely see it Bluetooth and Wi-Fi. There's also a Bluetooth and Wi-Fi chip on there That you can use for you know Wi-Fi or wireless connectivity again the default example It doesn't do anything, but you the hardware is there for your use When you press a button it will on the board. It's needed. It's all in one It will register your face So it says okay registering and you have to shift your face around a little bit So it gets all these angles of your face and then it says okay You have been added it knows who you are and then you can save That to memory if you want for long-term use and you can save The number of faces you can save is a couple hundred or to a thousand It depends on how precise you want it to be I think it's probably a memory or storage or just search speed thing It's like the faster you want it to go the less faces it has to be able to look through So if you only have like a you know a hundred faces it has to recognize You can use a slightly stricter model than if you're having to measure a thousand faces Did you have a program that can let you do a thousand? But again, you know, it doesn't it's not as it's strict So you you have to make that decision in the data sheet they talk about these these choices Once it recognizes you later it can you can save it as a name You can say this is the person's name or you just say user number one and then this sort of over thing overlay thing comes up and says welcome home and Makes us you can't see anything else So that's the basic demo. So it works quite well I tested it and like I kind of like, you know, put my hair up and my hair down I tried like wearing a hat. I tried different lighting and did a pretty good job One thing that's interesting is that when you press a button it goes into Infrared mode so it does have two cameras and I from looking at it It you know when one of them has a slightly different lens than the other I believe that one of them is IR and one of them is RGB and so it has Stereoscopic vision, but one of the nice side effects is it's kind of like an anti spoofing technology So one of the problems with these simple non-depth based Camera face detectors is that you can often trick them by taking a photo of a person and holding up that photo And the cameras like ding dong, you know the person. Okay, that's them. I I agree But when you have IR and I think if you have these two cameras that are offset a little bit that probably helps, too When you hold up a picture of a person In color and then in IR It doesn't look right. You know what I mean? Like the the flatness the infrared doesn't come out the way it should on a human face because The background and the image is flat And it's not like depth detection, but like humans radiate a different infrared pattern on their faces than a photo does and so When you open up the serial console, it's cool. It says IR fake face. It says like I detected a face But it did not pass my fake face test So it it will not get tricked by a piece of paper or a TFT with a face on it Which is you know kind of handy. I think that was it a nice you know, it's always a little embarrassing when you see folks like do like security analyses of like Like lockpicks or you know face recognition or fingerprint and it's like well, you could trick it with like a piece of Jell-O That's like here's a picture of Jell-O. Here's a picture of Jell-O and it's a hi welcome home Okay, but it doesn't have that problem So here's the algorithm So it can use deep sleep mode, which I thought was really useful one of the Nice things about microcontrollers is and you know Linux is more powerful Of course, there's there's face recognition software and open CV and all that good stuff But like suddenly you're running something that's going to draw about an amp You need You know a lot of complex software to run it to boot it up takes a long time. There's a chance of file system corruption What's nice about these? Stand-alone microcontroller Dev board versions of face recognition is one you don't have to worry about your data going to cloud or some somebody hacking into your firmware just because like there's a you know a backdoor password But it also can use something like the built-in PIR sensor It doesn't start doing the face recognition or recording any video until somebody is in front of it Which is like a kind of obvious thing But because the microcontroller can wake up within like a hundred milliseconds of the PIR sensor to a human It's nearly instantaneous like they don't notice the difference because it happens so fast Whereas if it was a Linux board, it would be like, oh, hey wait I got a boot up. Hold on like f-disk got to check out like all my sound cards, whatever a minute later You're maybe booted so Fast wake up, you know low power usage. That's why you'd want to use this instead of open CV Okay So they've got a nice web page that I checked out. It's got all of the details. It was super quick to get started so the this board itself is the The SLN business IOT is the dev board or an it's ready to go It has everything you need to do like your project or product you can design with it And then of course you can buy the individual chip as well So just watch out like I'm we're talking about the dev board, but they also sell the chip and chip is like $7 It comes with the license and software and everything for this. It's a BGA chip NXP will help you with getting it integrated into your product But I would get the dev kit because well friend of you could probably just use this as is it's a great edge, you know facial detection system and they have The code you just need like a J-Link programmer to program it you program it up with NXP NXP Expresso I think is the name of the ID But it's all open source and available like the source code for the demo so you can hack and mod it And of course like I said, there's a TFT screen as well Okay available on digikey site pick it up the folks who want to know the link you can do digikey.com 4-slash short 4-slash 4 7 5 V 9 4 or you could probably just search for what? Visnas honestly the viz and as nothing else name that and then if you want to check out There's a board a show video. There's a bunch of videos also on the NXP site showing the different things There's also a couple other details. I wanted to mention real fast It also has an emotion engine That you can you that you can turn on that will detect like happiness or surprise your anger I don't know how useful that is unless maybe you wanted to know if someone was angry at your door And he wouldn't let them in I don't know but it does have that capability There's a command line interface that you can go and you can like turn on like debugging an interface Get happy before you come in it's kind of like wiping off your feet now Come in with a smile on your face. All right. Do you want me to show this on the overhead? Yeah, we'll show it on the overhead and then we're gonna play a video for NXP Yeah, should the video and how you get it registered. So this is the dev board So this is the camera interface and you can see there's two white LED so you can turn on the LEDs to illuminate somebody's face And then there's two IR LEDs here and the two cameras and like you can kind of see one has a slightly different filter on the lens I think so one of these is probably the one with the IR cut filter and maybe this one isn't or I don't know And then this is like a like a milled out piece of plastic. This is the PIR sensor USB four buttons This is probably where you would connect the TFT if you want it and maybe some external hardware here This is the dev board module, which I think is just adorable. This is the IMX RT 10 6 f f is probably for face windbun flash or Yeah, ISSI Ram this is the BLE and Wi-Fi Chipset probably SDIO and then it goes to this like antenna, which is on the back here So you can do bluetooth or Wi-Fi control and then this Plugs and I didn't unplug it because I don't want to break it. But yeah, they got all the passes on the back Big power supply inductors, maybe a PMIC and some other just management chips more power supply stuff And then yeah, this is this is the this is the microcontroller board and Wi-Fi So this is like kind of cute nice plugs into the user interface board and then this is the camera board which is really Stuck in there, but it looks like this. So this is what you've got here. I like this dual camera. It's kind of a cool Cool thing. I've never seen before but there you go. So these three pieces and they plug into this milled resin Package, okay, so let's let's watch this video showing the process of registering and identifying names I'm kuber karnahan a solutions integrator here at nxp And today i'm going to be showing you how to get up and running with your sln vision as iot kit When you first receive your kit you'll find three things inside a quick start guide a usbc cable And most importantly your sln vision as iot kit To begin take the included usb a to usb c cable and plug the usb an Into your computer's usb port and the usb c end into the usb c port on your kit The usb port on your kit provides power to the board drag and drop flashing via msd And virtual serial port capabilities with the cable plugged in a green led will light up to indicate the kit is receiving power Once the application is running Your computer will automatically discover the kit is both a calm device And usb webcam and your computer may make a sound indicating it is detected a new usb device Next open the camera app if using windows or cheese If using ubuntu to view the camera output coming from the kit if you have multiple cameras attached to your computer You may have to switch to the sln vision as iot camera to see anything coming from the kit Right off the bat you'll notice the built-in GUI provides some convenient on-screen information Including a bounding box to help users properly align center their faces while registering The on-off status of both the wi-fi and vle capabilities of the kit And the total number of users registered in the kit's local face database Press the switch for push button on the front of the kit to begin registering a new face An on-screen message at the top of the screen will be displayed to indicate that registration is taking place Simply hold your face in front of the camera making sure to center your face within the provided guidelines Upon successful registration a message will be displayed indicating a new user has been added to the kit's local face database And the number of registered users will be increased by one If registration fails simply press the registration button again to retry Once registered your face will prompt a welcome home message whenever your face is recognized Registered faces will continue to be recognized until the kit is powered off or the face is manually removed Although saving faces permanently can be done using the cli or by enabling low power mode Both of these features will be discussed in more depth in later videos This concludes our look at powering on and registering a face with the sln vision as iot kit Check out the next video in our series to learn how to test out the liveness detection and anti spoofing features of the kit Okay, watch all the videos because yeah more videos when we show you the first one this week's ion and pi Check it out on digikey