LIVE Shop Talk 10: Review of Jetson Nano After 1 Month of Use.

[Applause] you hello guys this is pulmicort from top tech boy comm and we're here today with our live shop Talk number 10 where we are gonna be talking about my review of the Jetson Nano after using it for one month and I think maybe the intro has queued you in a little bit as far as what my experience has been and what I wanted to do was just kind of talk to you guys a little bit today about what it's been like playing around with this thing for about a month and so hope you guys will ask some questions and we'll kind of catch up with some other business on this on this channel but I will say the bottom line is trying to get this Nano to do what I want has been a rather frustrating experience but it's so exciting and it is so incredible the power that it has and what it does you just don't want to give up and so it has been a fairly intense a fairly intense month as I've been trying to get this thing figured out so I will need you guys to pull your poor yourself a nice big mug of iced coffee I believe down in the chats I saw some some rebels out there drinking iced tea well teas ok as long as you don't put sugar in it ok iced tea iced coffee either one is acceptable and get ready because we're gonna be talking about the jet the the jets and nano today and a little bit about what my experience has been and I will say first of all that it's been probably one of the most challenging projects that I've ever done and the reason that I think it's so challenging is if you guys get an Arduino and I have an Arduino I'm trying to think usually I've got Arduino laying around on my or do we know laying around on my desk I don't see one right now but if you have an Arduino and I have an Arduino we pretty much have the same Arduino right so if I show you a command and you type in the command then that command is gonna work but it's a little bit different with the jetson Nano and the thing is is that the Jetson Nano has some very very highly specialized PAC in software and things get a little bit more confusing because it's we talked about in our earlier review of the of the Jetson Nano it has a hundred and twenty eight cores and so it's like a hundred and twenty eight micro processors that are all operating in parallel and for you to exploit the power of those parallel processors you have to have very specific software and so to utilize the power of those processors on the Jetson Nano you have to have a very specific version of Python running with a very specific version of numpy running with a very specific version of open CV and tensorflow in all these different packages and so kind of like the nice thing is is that when you download the image card for the jetson nano from the Nvidia weds website they have all the right packages put together in a package and then you download it on to onto your computer and so then it it has all the right versions but then if you're fooling around with it and start doing something very quickly you get it configured in a way that it doesn't work right and so what it's the first thing that I've learned with this is is that you've got to be very very careful and very cognizant of the version of the software that you're using and you don't just go in there and randomly upgrade from one version of OpenCV to another version of OpenCV another problem that I really had was that you know I'm kind of like comfortable with Python and so I want to do my development on the Judson Nano with Python and then right off the bat there's not on there like a Python integrated development environment and you're using just a text editor well if you get used to writing Python code you want a real development environment that'll highlight your text and help you fill things out and stuff like that and so first thing I wanted to do was install an integrated development environment for Python and what I found is I could install it but then I could not get that environment to work with the packages came on the Nano and that if I had tried to install those packages under my development environment they weren't the right ones to work with the Nano and so I spent a good part of the last month just trying to get the Nano with a good Python editor that I could write code in okay the second thing that I found is is that this is this is really really a lot of work to get this thing to get this thing going and so one of the things is like if you go to the tutorials there's just a lot of type this command this command this command this command this command and a lot of you typing commands that they're feeding you and you really not understanding all of this stuff of what are you doing what does the command mean what if I wanted to do it myself so they're giving you commands that work but they're not they're not explaining it so after working on it for quite a quite a bit of time I finally got all the demos up and working where I could do the facial recognition and the image recognition all that stuff and then man that starts getting you so excited when you can see what the Nano can really do you start getting vetted and you start to want to learn more okay you start getting very excited and you want to learn more so I kind of got to that point that I had the reviews working or the the sample programs working and that really really really got me excited where I wanted to learn more and one of the things that makes me so excited about the jets and nano I will show you the thing here is you know it does all the things that you think about with artificial intelligence and all the things that you think about with machine learning and all of those really exciting things but what's different about the jetson Nano is it actually has the GPIO pins like if you think about the the digital right in analog ride and digital Reed peds pins on an Arduino that you can write to or read from those are on the Jetson Nano and so what that means is you can do your machine learning you can do your artificial intelligence but then you can use that to make decisions in make a motor mover or make a servo move or be incorporating data from sensors and so you having a way to not just have artificial intelligence but to gather information from sensors in the real world and then affect change on the real world through stepper motors or motors or our servos or things like that and that's what just has me so excited about this thing it just very very exciting it has motivated me not to give up and a little bit of the frustration with the demos that come with it in the demos that Nvidia put together they're really for someone that already has a PhD in machine learning they're not something that you can really learn from so I say oh wow look at that that's the coolest demo I've ever seen but I don't have any ability to change that I don't have any ability to make it mine I don't have any ability to incorporate my own ideas into it now what I will say is I think I sort of cracked the nut and I actually am to the point that I am able to start developing things on my own and the good news about that is if I can do it then I can teach you and so I would say that I am getting very very very excited about putting together a tutorial series on this that I think I can step you through and I think that I can teach it to you because I kind of had that breakthrough moment where I now have a Python environment installed and I now can use libraries that I can bring into Python and then I can start writing programs that do what I want it to do and not just copy someone else's program so that's why I put the clouds I just feel like I've really reached that breakthrough that breakthrough moment so let's see let's see what comments you guys are making Joe coffee tip you might like I got refillable pods for the carrot yeah I am really looking into that some of you guys have been sending me coffee pods for the Keurig little coffee pods thanks really appreciate it I appreciate you guys that are supporting me over on patreon - that really helps me keep this stuff going and you know there's a lot of gear I'm buying I really need some new computer equipment and so appreciate you guys that are helping me out on patreon look down in the link you can go over there look down in the description there's a link to my patreon account and also I want to kind of start talking to you guys about whether there's some interest in me doing a series on the Jets and nano and it's sort of like artificial intelligence for the complete beginner what I want to do is I want to take a guy that is smart and hardworking and wants to learn but doesn't know anything about artificial intelligence and take you through it step by step so you're not just wandering around in the wilderness like I spent the last month that's kind of what I'm interested in yeah so I would like to get a Keurig right now I would really like to get the curing system I would like to get the little pods yep I think you are and I are think I'm thinking the same Joe you can even put tea in them hey that sounds good okay Gary how are you doing Thomas hi act shot let's do this yes sir Prince and more hi okay a bunch of you guys checking in hello from Massachusetts with coffee okay very good can you turn off the flashing background oh you didn't like my lightning in the background well man that is my month for the last that was my month was going through that wilderness going through that storm for you guys to try to figure out all this stuff okay man if you're not getting notifications on your subscribe go to my channel on the subscribe to make sure that you hit the little bell and then you will get notifications did you install the Nvidia SDK yes okay but there's different Nvidia SDKs and what we're going to have to do is we're going to have to make sure that we're operating on the same one so I've had like two of you order a Jetson nano I want to see are there more of you willing to get into this the basic Jetson Nano is a hundred bucks if you get the tricked out system the tricked out system that I have here it was a couple hundred bucks but I bought you know the little wireless keyboard i bought the got the Wi-Fi module the little camera two hundred bucks you get all the bells and whistles probably you're gonna need at least a hundred and fifty because you're going to need nan-oh you really really really really need the fan you need a good power supply and you need some sort of camera you could use a webcam you can see that I'm using the version two of the Raspberry Pi camera you could get by just with a webcam if you already have one a little USB webcam but are you guys up for this right are you up for learning artificial intelligence are you up for learning this Jetson nano if enough U of you are interested I will begin to put a series together like I say there's been a couple of guys that have ordered one and I don't want to leave them hanging but I need to see some other people and the focus continues to vex me okay let's see yeah I do use the NVIDIA SDK that was a question Keith says if if you want to do Python io on the Nano you can try blinka from Adafruit it's a very nice package yeah but again you know you're kind of then like you've got the Jetson nano development going this way and now you're off here on a side branch and so we've just got to be careful and we got to make sure that we're going down a path that we can get to the end and that's the thing that I'm kind of a a little bit afraid of but I'm going to show you some stuff that I've done in a minute that I coded up from scratch on my own not using someone else's demo program and I'll show you that in a minute trying to get some of your question compiled for Nano the run the app okay there are people that are building their own environments and then compiling it for the Nano and building I'll tell you that's beyond what I really feel comfortable doing so I want to get a core install from Nvidia and then I want to develop on that platform and that's kind of the direction that I am going let's see I've got Shaquille hey Paul big fan of yours great I hope you keep churning out those tasty tech videos okay I hope so too I hope so too you guys give me some encouragement there would you start any series of machine learning tutorials that's what we're talking about that sort of like what I envision is AI for the absolute beginner based on the Jetson Nano and then we start developing projects where we're interact with a real world and rather than just doing abstract things on the computer actually start kind of building real projects okay that's what I want to do hey Andy good to see you just monitoring here we'll go back to Arduino tutorial way above my level al al how are you doing mr. al ludic ii yes the the most excellent al Luda key okay my nano project is an autonomous push lawnmower oh man be careful about that getting power equipment automated that could there in a lot of ways that can go wrong pi Image Search has done some great articles yeah they have and I've kind of been watching what they're doing he gets to the point though and he doesn't install OpenCV and the problem for me is is that I see again he's like a brilliant crazy insane genius like the problem is though he's like so far out there smart am I really able to learn from him and one of the problems that I see with what he did was he kind of stopped and didn't do OpenCV and if I'm gonna teach someone who's new I can teach you to write programs in OpenCV okay and then you can actually be programming real artificial intelligence stuff and then gradually move into the more complicated stuff but I see open CV is the path to actually teaching you and he kind of kind of skipped that but I will tip my hat pie image search guy is way smarter than I would ever hope to be okay interested in learning a karoon good I think 100 euros is a bit much for me so maybe I will consider cheaper alternatives okay that makes sense that makes sense okay ah Mike Rubino has the Jetson Nano okay well we need to we need to start thinking about what what to do with it if I did a series the first thing I would do is I would show you how to download and install the software and just basically get to the point that you're booting the desktop let's see if I can show that here okay so you see I have a desktop booted up here you're actually looking at the real thing here you are looking at the Jetson Nano and you can see it looks like a lot like a Windows desktop you can see here we could go to most excellent ww tech boy calm okay and we can call up that website and then we can come over here and we can open up a spreadsheet we can open up a word processor we can open up something that's like PowerPoint and so you see you've got all the things that you sort of would think of in a Windows machine of course this is a boon to desktop but you could do all the things that you would normally sort of associate with a with a desktop computer so I would get you installed where you had an environment that we could start developing on then what I would do is and if you guys already know Linux don't hate me but remember we need to like what's our motto leave no man behind or leave no woman behind don't leave anyone behind so the first thing that I would do is we want to kind of get out of this Windows environment and we want to start operating in a terminal where we would be learning the Linux Linux commands like LS show me it shows me what is in my directory I could go see Dee and slash that takes me to my home my root directory LS what shows me what's there I can do change directory into the squiggly which takes me back to my home directory which I can list see see we got to get good at kind of navigating in the terminal so you're not sitting there with those doofus folders and that goof that goofy you a graphical user interface but we're operating from the terminal because there's a lot of stuff that you just can't do from the folders so we got to learn how to work our way around the Linux system so that is the first thing that I would teach you and the next thing that I would teach you is first of all install the Python environment that I am using because this was my biggest problem was to get a good Python environment which you can see that I have here and you see how it nicely formats the code it highlights things that understands it helps you debug it's a really good environment but it's working with that Python and library install that comes with NVIDIA SDK so you're sitting there in your developing code on that environment that Nvidia put together and they're like a lot smarter than I am they're way smarter than I'm ever gonna be so I need to work in their environment but I need some code that will allow I need some environment they will allow me to start writing coding so this is a program that I wrote okay you can see that maybe it's 30 40 lines I hats let's see let's actually look it looks like it's 87 lines long but a lot of that is comments but you can see that I'm importing libraries and then I'm setting my camera up and then I am setting some variables and then I have a while loop and then I'm doing this stuff so so you can see that this is kind of like just a simple Python program so I would teach you how to install the software I would teach you a little bit of Linux I would teach you simple Python like maybe five or six lessons on Python and then we would jump in and we would start doing some of this machine learning okay let's see do you guys have any let's see do you have any comments here okay mike has the macro Vino has the Jetson Nano does OpenCV be used only on Nano no Nano OpenCV will run on anything the hard thing is to have a Python environment on the Nano with the right OpenCV so that you're taking advantage of the processing power those 128 processors because if you just dump LIGO in and install OpenCV it's just gonna run it on a CPU it's not going to be running it on that parallel mode how many times faster is a Jetson than a standard current eight core PC film factory that's kind of hard to answer because it's like how effectively like first of all if you're running Windows on anything Windows would make anything slow okay and then how intelligent is the software that you're using to to utilize the eight cores that you have the thing about the Nano is you've got a hundred and twenty eight cores and then you have a really slim efficient operating system of the boon to Linux okay and then you've got libraries that are designed to go in and utilize those 128 cores and also it doesn't matter how much core processing power you have if you don't have the paths in in the past out and that's something that I think Jetson Nano did a really really good job at is giving you really large pipes really large and accessible big pipes in and big pipes out and so this thing really really gives you a whole lot of processing power okay if you get stuck let's see DM Paul if you get stuck on any aspect okay I'm not exactly sure there hi hey Charles how you doing welcome okay would you recommend getting Jetson instead of the Raspberry Pi that's a good question okay I just ordered the Raspberry Pi 4 and I've got the I've got the Raspberry Pi 3s and man I just don't quite get the Raspberry Pi it's like it's kind of like I'm not sure what they want to be it's like I don't want a desktop computer that sits there and has 50 cables coming out of it it just like is it a desktop computer if you try to use it in an embedded application it's really easy to corrupt the card like simple things like an on/off switch you know things like that aren't there and so I ordered the Raspberry Pi 4 but it's like what for like maybe if you're interested in gaming or retro gaming and you make a little gaming console or maybe if you're interested in an entertainment center where you're like streaming video or something okay maybe I see it but I didn't know really what to do with the PI 2 and so with a PI 3 I don't know what to do even faster than it was with the 2 and now I've got the Raspberry Pi 4 and like really really really very very fast I'm just sitting spinning my wheels because I don't know what to do with it now the Nano I get ok I get it and I see where I'm going with it like I have all these crazy ideas in my mind of what I'm gonna do with them with the Nano it's like artificial intelligence interacting with the real world at the local level it's like some really exciting things happen the Arduino really exciting things going on there I know where I'm going I know what I can do I don't quite get the Raspberry Pi now I do kind of like the pi zero which I think there's some cool things that I reached the point that I really don't know where I'm going with with I run out of horsepower on the on the Arduino and therefore I can kind of see a place for the pi zero but I really I'm kind of like I'm making up my mind my future direction on some of these advanced projects is gonna be with the Nano and I've already got some really really good ones in in mind okay would you write yes I would get the Jetson if you guys are interested down in my description there's links to the gear that I have on the Nano and you guys might want to think about starting to put your gear together there okay I'm up for that I want to learn those things okay good start getting your gear put together please suggest any good torque tutorials on open CV from scratch there are no good ones the problem is on open CV they assume you already have your PhD so that's what I am going to do I'm going to teach you how to boot the the Nano I'm going to teach you how to get it configured I'm going to teach you some Linux I'm going to teach you Python and then I'm going to teach you open CV now if you already know some of that stuff that's okay but we're gonna start from absolute zero and we're gonna build up and then I'm going to show you some cool things that you can do with open CV I'm planning to install Linux Mint on my old laptop with windows 7 bites the dust maybe I should just do a Jetson nano for my desktop competing needs instead well you know the Jetson is not really a laptop machine I mean putting putting Linux on an old laptop I think that's a great idea in fact I sometimes wonder why am i keeping windows on this computer this broadcast software that I'm used has to have windows so that kind of keeps me on Windows but man this today's coming that I'm just gonna go completely to Linux okay let's see this is my program that I ran from scratch do you want me to run it class class you want to see what I did you're gonna see my demo anybody out there does anybody care okay you guys don't want to see my demo yeah I know there's a 30 second lag so I've gotta wait let's see I'm laud Anderson from Ghana hey LOD how's it going how are things in Ghana I heard was there like a bank run there recently or was that a different nearby country I just heard there was some difficulty in Ghana I've never been to Ghana but I would like to visit there okay thank you very much laud love it over there in Africa I love to start from scratch and learn the fundamentals okay yes yes okay that's my cue you want to see my demo okay I'm coming over here can you see my mouse now you see look at this nice environment I have this is really nice and you see I can have different programs up here under different tabs okay and now I'm gonna right mouse click and I'm going to say run from the Python terminal okay and so I'm gonna run it okay and then look at that alright so you see me up here alright and then what do you see you see a big black empty screen and then what you see a big black empty screen and then what do you see you see another big black empty screen so you see me live and then you're seeing all of these different screens that don't have anything on them in fact so many that we're not going to be able to look at them all okay so what is this this is a webcam okay and it looks like that in the overall system there's a couple of seconds of latency but you can got to remember with this broadcast stuff it's going a lot of different crazy directions but I just see my live camera on the broadcast is maybe a second or two off from the screencast on the Nano but let you see one two no that's like maybe what do you think a half second latency one two maybe a half second leg latency so that's not bad so this was just starting from scratch I wrote a program that would fire up a USB camera and and then would start doing image manipulation not very interesting right now but watch this okay watch this watch this now okay you see I wrote a program that would recognize and find and track okay so now all those screens that were Brent Blank why were they blank because there wasn't a red pin in there there wasn't a red pin in there but when the red pin comes in it finds it and then do you see in the live image how it puts a box around it okay now the other ones are doing different filtering to try to get the most perfect selection the most perfect finding of the box that you can and so I think that that's pretty cool now again I hope you understand I didn't just run someone else's program or copy someone's program I got it configured I got my environment I wrote a program and so now I can do all types of things also that one where you see that it's white do you see that one where everything is white that's creating the mask I create the mask then I mask everything except the pin and then I put that image back into the original image and then I add the blue box around it do you see how that's tracking look at that uh-huh boom from scratch from scratch I think that that is pretty cool okay okay Majid from Ghana all right man it's great to uh it's great to see you okay are those mapping options not sure what you're saying on mapping options yeah that you went boom are you bet I did the happy dance right I did the happy dance when I got it working but once I got my Python environment working I really knew that I could do it then I mean once you turned me loose on a good Python environment I was sure that I could do it and I started writing code and so I'm doing simple things that could recognize things okay now like I want you to think if I do you see that box do you see that box there what does that mean that means in the code let's see if we can come back to the code here real quick what that means is in the code where is my environment okay that means in the code I know where you know x and y are I know where with an x value and a y value look at that so I know in X and I know in Y I know the x value in the Y value of this pen tip okay if I know the X and the y value of this pen tip in real space what does that mean well since I've got GPIO pins I could have like servos where I could have something that you know like imagine this is what I'm imagining this is the target and this thing would point and I would have a physical object pointing at this object in the real world because of the Nano GPIO pins and that's like really really exciting to me that I could do something like that because I thought could point at it then think of all the interesting things that you could put on the servo and then start thinking about all the crazy projects that you could do and so I am really excited because I've demonstrated enough to know that I could do very very interesting projects here so let's come back over here okay let me see do we have any more questions okay okay Andy okay yeah you looking good sir thank you ochre loved a Richmond thank you for saying I'm looking good feeling good man I am feeling really good I am really really excited about this again I'll show you here's here's where we are no more thunder and lightning right we have the happy little clouds this is how you feel when you get something working and so that made me feel really really good and hey are you guys tell me are you guys also on the are you guys tuning in to the to the nan axis it looked like we got a lot of people on that project are you guys following the 9 axis because kind of like one of the exciting things to me is I really see that possibly I see that non-access project possibly really playing in with this because one of the things that you could use for input to the AI system would be that those quaternions and then you know there's just all types of all types of things that start coming to my mind that I could do if I could if I could do that and so I think the non axis is going to be a player in these future projects as well okay and then like if I come back to my windows let's see if I can get my windows back up I do not want that let's see okay here's my window right and this is the one that's tracking okay so to quit I just click on this and I type Q all right and the thing quits let's look at this other one here and this one what I did is instead of a webcam actually have the pike am working on the Nano and so let's see if we can make that one run that my friend was not correct somehow it ran the same program let's come back here to this one right mouse click and say run Python file in terminal okay see see this my friend is the PI cam and you can see that this is the camera that I have looking down on the PI so this is my system looking at the PI okay and then this is the PI looking back at that but what we can probably do is turn it over here and now I'm kind of in the view but that's that little PI camp so to me that's really exciting that you got this tiny little postage stamp camera and that I'm actually interacting with it and can it find the pin but look at that I'm finding the pin with the finding the pin with the with the PI cam and so man what's really exciting to me about OpenCV is what you can do is in OpenCV what you can do is you're streaming data from the you're streaming data from the camera but you pull it in one frame at a time and then you manipulate the frame and then you show it so it's sort of like you're sitting there between the camera in the display and that's where the magic happens and so I am really really having a lot of fun with that that is very very good okay I love your videos very good Ameen thank you I mean very good to see you Charl yes can't wait to get my paws on the 9 axis okay yeah let's see Charl did you order it did you say it was like you ordered it and it was in customs or was that someone else I can't remember if you had it or not very interesting in the non axis okay you guys I got kind of a lot of hate on that because I just sort of showed the introduction and I think people wanted to start coding Thursday tomorrow the next lesson we'll wire that thing up and get it going okay why had multiple windows with the pin okay that's a good question and let me try to answer it what I'm doing is I'm doing different image processing okay so let's just start here okay let's start here and the thing that you're seeing with me and it that's the end that's the composite thing where I put everything but the first thing is I have to find the pin and so that is that black with white right that is that this one right here I hope you can see that that's this one and then though do you see over here can you guys see all this noise these little white dots that shouldn't be there I've got to get rid of that so I come in and like this was a closing mass cry try to close off certain thing you can still see I have errors there right I still got errors and then I do different masking this one this opening mask look at that you see you see it finding it it's finding it well and I've gotten rid of most of the noise but like all of these I probably would not have all of these different things but I'm trying different filtering techniques to have the best sharpest understanding of where I get lost where the camera is of where that thing is and not having any false signals okay and so I'm comparing different filters but in the end application I would just use one one filter okay so Charla's non axis well that is great soon you'll be ready for that lesson tomorrow and we will be hooking that thing up okay we will indeed be hooking that thing up let's see what happened there let me try this one more time okay let's see what happens I'm again I'm trying to manage a lot of different things here all at the same time so me let me come over here I'll get back to your questions in just a second and the other thing is when I'm using the Nano and the PC at the same time and there's two different box two different mice invariably I grab grab the wrong Mouse in the wrong keyboard so again guys I hope you guys are gonna I hope you guys are gonna join us on this side remember this remember this right let's see I gotta flip this thing around here so we're in there it is that's pretty good look at that okay so this is the non axis project and this thing is going to really kick off in earnest tomorrow seems like there's some really good interests I've had quite a few people I've had quite a few people that have expressed interest and I think number of you have your gear already and so that'll be that'll be good so this this is back over here to the Nano and this project let's see if you guys have any more questions here let's see where we are on questions okay or am i bus address okay I'm not sure what the question is on the my bus address are you filtering for red or the shape okay film factory you're getting into the nuts and boats I'm finding red okay and that was just the easiest way to get started because I didn't want to do a bunch of training I just wanted to see can I find X&Y and can I interact with it in color was the easiest way to do it so you see if I put something else for as my camera if I put something else up here that's red it will track it but you see I have really good delineation if I put things that aren't red and it's a pretty bright red that works it's a pretty specific color of red that actually works I could also do shape and then beyond that then you start doing your training and you start having your Cascades okay but if I can do this then it's just a matter of instead of filtering on color filtering on a cascade then it gets into building cascades but you see you have enough success you have enough success that you kind of get in this happy mode that motivates you you see you get in the happy mode that it motivates you to go on I swear I was to the point that I was thinking I'm never gonna learn this and I'm never gonna be able to teach it to anybody and I better just quit now but yes I will I will do that okay Charlie p'suer but still way less than most people spend in a bar in one night oh yeah man or like what people spend on video games or what people spend on nonsense it's just like this is a little more pricey than other platforms but man those of us with electronics and programming is a hobby we spend a lot less than people who go bar-hopping or you know play golf or have a motorboat or something like that okay do you save the red tip image then compare it no no what I do is I adjust the parameters let's come back over here and let me show you again way too many mice here to keep track of let's look in the code and let me kind of show you a little bit over here in the code view what we're doing can you guys read this do I need to make my font bigger are you able to read this okay so let's see it in this code up here what I am doing is you know how colors are RGB for red green blue well for this it's easier use HSV hue saturation and value and then what I am doing is I am playing around with my hue saturation and value numbers okay do you see here I am setting ranges of hue saturation and value and I do that such that it always gets the red but it doesn't get other Reds so I want it wide enough that it will always kneel it but not so wide that it starts getting noise in the background so I sort of define this and I did it just with playing around but the way we're gonna do it is we're gonna make slider bars and we're gonna put the object of interest and then we're gonna slide those slider bars until it just absolutely nails it so tweaking these parameters is part of what a lot of part of what we're doing okay a cascade is this enormous data set that like imagine that you took a thousand images of what you're interested in like a thousand images of watches and then you had a thousand images of things that aren't watches and then you just turn this huge processing power loose on it where it just turns and turns and turns and turns and turns and it learns what a watch is and it learns what a watch isn't and then when it sees a scene it sees if it sees something that it infers is a watch but the reason it's hard is like look at this it could be this you see like think of all the different angles that the watch could be and then think of all the different things that are watches and then think of things that are not watches that are that are maybe similar to a watch like maybe a digital thermometer or something like that so the the cascade is affect the output of the training because you couldn't train in real time because it's too computationally intensive so you've got to create the image database you have to train it on that database and then you have to use the sort of trained model okay I wonder if you can use a Wi-Fi or ham sensor instead of a camera to automatically track that would be interesting again we got the GPIO pins and then we've got the horsepower so you can use different sensors you can do the computations and then different outputs yeah I mean I'll go through this I'll go through this with you this isn't ready for you to download yet because it's just done but if we do lessons then I will do it okay I will do it all right let's see is this capable of learning that red pin just by showing it the pin or it must be defined in your code well you see that's the whole thing like like right now I'm just training it by picking the color and having it match the color the next thing is we could teach it to recognise the pin based on the overall parameters which would be the shape and the color in the riding and all those things you could train it based on that and then you could train it like more broadly where let's say that you could show it and it would say that's a red pin that's a blue pin that's a green pin so there's all different ways to train it but I want to get to the point that we're doing something because you'll lose interest and just give up all hope before you start getting results so we've got to have it where we're getting those little happy results as we go along okay all right so if you guys are interested some of you guys think if I see you know again links down below for the the gear that I have if you want to play along you've at least got to have the Nano you have to either have a USB camera or the paw cam you really need the the case the case is sixteen bucks that includes the fan so that's not bad and then you need you need a good power supply cuz this thing draws four amps you need a good power supply you could use your own keyboard you could use your own mouse you could use your own cables you could use all those things yourself but you really need the case with the fan you need some sort of camera and you need the the Jets and you know I would get a nice I've broke down and just I got those little mini SD cards all over the place but I got a big new fast one just so that you know if you're dealing with something like this you don't want to be slowed down by a by a doofus doofus memory card okay this has been pretty fun you guys we're going on 43 minutes here and I need to get back to work so I'm gonna let you guys go you guys tune in tomorrow we'll do a will do another non-access tomorrow Tuesday coming Tuesday we'll do another of the first series of Arduino and I just start needing to get feedback from you guys whether it would be worth me doing a lesson on this on this nano or not I look on YouTube and the lessons that are out there on artificial intelligence in the nano they don't get very many views almost no one's doing it but then the question is is it that nobody's doing it because it's too hard and the guys are kind of like not real friendly like I asked someone a question on the forums the other day and they didn't answer my question they sort of kind of chided me and with the energy that they expended in chatting me they could have at least pointed me in the right direction and so sometimes these guys are so smart they don't have time to you know they don't have time to waste on people like us that aren't you know already PhD artificial intelligent computer scientist and so maybe it would be more popular if there's someone who was a little friendlier and go a little slower that would get them up to speed so give me some feedback if that would be something you guys would be interested in if I did a series on artificial intelligence on the nano with there people following along bottom line is it was very hard for me to learn it it was very very hard for me to learn it to get started but once I started having the breakthrough now I'm just extreme I did and I can take you guys step through step by step through it so you won't have to go through all the pain I did so give me some feedback about whether you guys are gonna be interested in this or not you guys give me a thumbs up subscribe to the channel make sure that you've hit the little bells so you get notifications share this with other people let's see if we can get more people working on this stuff palma quarter from top tech boy calm I will talk to you guys later

Channel: Paul McWhorter

Views: 6,071

Keywords: Nanotchnology, Jetson Nano, AI, Opencv, Review

Id: 23hPle04BV0

Length: 47min 57sec (2877 seconds)

Published: Wed Aug 21 2019