Review of the New NVIDIA Jetson Orin Nano

hello guys this is Paul McCarter with toptechboy.com and we are here live live live live and are you guys ready to rumble in the Jetson jungle if you are ready to rumble I will need you to leave a comment in the chat ready to rumble because if you guys are not ready to rumble we can just end this live stream right right now because if you're not excited about this I'll just go back to my needlepoint Channel because those guys they know how to rumble we'll just forget this high-tech stuff but let's see if you guys all right Andrew is ready to rumble let's see if we've got a few more people Mike ready okay that sounds good and as you guys are responding I will need you to pour yourself a nice tall glass of ice cold coffee that would be straight up black coffee pour it over ice no sugar no sweeteners none needed and also what we're going to enjoy this morning as we are doing a review of The Incredible new Nvidia Jetson Oren Nano we're also going to be watching Dawn break over the headwaters of the marina river right behind me here so we got a lot of great stuff we got the Dawn breaking on the Nile River we've got the Jetson jungle and we've got our coffee what could be better than that well don't answer because some of you might answer inappropriately but it's pretty darn good to have those three things let's see who we ow that was loud let's see who we've got in the chat we've got Mike Andrew Big Bob boggy face we've got uh Rob Brian so we got a good crew that is showing up you guys check in let us know who you are where you are and if you are ready to rumble okay second question are you ready to see my Jetson jungle let me know if you're ready to see the Jetson jungle because if you're not interested I'll just go do something else but do you want to see the Jetson jungle give me some feedback here okay looks like we got yeah it looks like we got some guys that want to see it let me switch over here to this View and let me show you the Jetson jungle that I have going here you will notice that I have yes I have our old friend the Raspberry Pi I'll talk about that a little bit because I know a lot of you guys are taking my Raspberry Pi class and I'll kind of talk about this Jetson business in context of what we've been doing with the Raspberry Pi then you can see that I have two models of the original Jetson Nano I also have a two gigabyte Jetson Nano I should say all of these things are plugged in booted up and running and so I've got two Jetson Nano four gigabytes I've got a Jetson nano two gigabyte and then I have uh let's see I'm going to switch over here I'm going to switch to a different view here where maybe we can see this a little better then I also here have the Jetson Xavier NX and then I I have the all-new freshly minted Nvidia Jetson Oren Nano here and this is the one that we're going to be talking about today and then of course there is this bad boy which is the Jetson agx orange now today I will really be talking to you about the uh Jetson Oren Nano now I'll have to tell you that this is not going to be your typical unboxing and review video what is the typical unboxing and review video they take the new computer the new board out of the box and then they show the spec sheet from the company blindingly fast greatest thing since slat spread you know it's all great and then the company provides these uh these can programs written by Immortal experts in the world of computer technology and then the reviewer runs the program someone else wrote and sure enough it's the greatest thing since sliced bread it shows in the demos what you uh what is claimed on the spec sheet and then that's the review well I'm not going to do that and the reason is I'm not saying that that's deceptive it's just people that are really really smart program it and then they get out of the board the specs that they're claiming and it's all legitimate but to me the real question is not what these brilliant experts back at Nvidia can do but what US mere mortals can do with these boards and so what I'm going to do is I'm going to show you what a normal person a guy like you or I that's been working on the Arduino and the Raspberry Pi what can we get can we get this working can we start getting some exciting stuff out of it because the question is yeah you could run the demos I could run the demos but then what have we really accomplished if we cannot do things on our own does that make sense I hope it makes sense well let me take a little bit of just a second here and talk to you about the specs just so that you have kind of a comparison of the power of these different boards that I'm just going to radically oversimplify it so that you kind of see just roughly what the differences are between them and so first of all the Jets and Nano the Jetson Nano has four quad arm processors so four CPUs and then it has a GPU that has a 128 cores and so you could think of it as something like the Raspberry Pi with a huge GPU sitting below it now of course even the CPUs on the Jetson Nano are faster than the Raspberry Pi but then you add that GPU down below it and that is where the magic happens it's capable it has four gigabytes of memory and it's capable of 0.472 trillion operations per second so we're going to round it and say that on these original Jets and Nanos you can get a half trillion operations per second for comparison this is just a very crude comparison but the Raspberry Pi would do something like .003 trillion operations per second and so the original Jetson Nano would be something roughly like a hundred times more processing power than the Raspberry Pi now 100 times more processing power does not mean that a program just automatically runs a hundred times faster yeah if you just dumbly did things on the Nano it would run a lot faster than the Raspberry Pi but what you have to see is the whole thing about these uh these products like these Nvidia products is to do things in an intelligent way where you're pushing the work from the CPUs down into the gpus and if you're sitting up here operating on the CPUs you're not going to recognize that performance enhancement that's that's possible with these boards and so that's what we're really going to do is see that us as mere mortals can we begin to see in the programs that we write can we begin to see the power the processing power that comes from those G CPUs I hope that makes sense now earlier this is a product that is a couple of years old now this is the Jetson Xavier NX and I really love this board this was kind of like the board that I used uh let's see I'm getting a little buffering here foreign let's give it a second to come back so this Jetson Xavier in X what it is it has the Jetson Xavier in X it has the Jetson Xavier NX has 384 Cuda cores so roughly three times what the Jets and Nano does it has six arm processors eight gigabytes of memory and it's capable of 21 trillion operations per second the Jetson Xavier NX 21 trillion operations per second compared to the Jetson Nano at a half so what would we say that's like 40 times faster if I'm thinking about that right about 40 times faster and then this Jetson uh this this Jetson agx Oren that's the bad boy and that one has 2048 Cuda cores it has four CPU processors four core CPU and it's got between 32 and 64 gigabyte lots of memory and it is capable of 275 trillion operations per second but what we are talking about what we are talking about today is the new Nvidia Jetson Oren Nano and it's really kind of in a sweet spot on costs that the Jets and Nanos are becoming available again it looks like they're coming in about a hundred and fifty dollars right now initially you could get one for a hundred dollars then they weren't available for a long time now they're coming back on the market at about 150 dollars pretty sweet the Jetson uh the Jetson agx comes in at about 2 000. well the Oren it looks like you're going to be able to get it for about 500 bucks so a lot of you guys maybe you'd just like to get a Jetson Nano but some of you old guys with a little more coin sitting around this is a really cool board and I really want to show you what it is going to be capable of today so again on the Jetson Oren it has 1024 Cuda cores in the GPU it has six arm core CPU it's got It's got eight gigabytes of memory 40 trillion operations per second if you can do it okay the processing power is there but are we going to be able to get down there and do it and that is what I am going to be looking into to today so again a little bit of context a lot of you guys have been working a lot of you guys have been working on my Raspberry Pi class and the thing is we have developed some really good skills using the raspberry pi but what happens as we begin to develop those skills what happens is we start developing more skill where we just run out of oomph on the Raspberry Pi and so I'm going to run one quick program which is something that we are doing in the Raspberry Pi class right now and sort of show you that we are getting pretty advanced in what we can do on the Raspberry Pi but the then we run into a little bit of a roadblock so let's see I think you guys can see this and uh it looks like I might be getting a little bit of latency guys it is raining outside today and you know what that means when it rains in East Africa internet doesn't work very well but we're going to try to power on through and I'm sorry that this morning I'm having not the smoothest uh the smoothest live stream but here we are on the Raspberry Pi and I'm going to come up and I am going to run this program and let's see what happens here okay so what you see is here I am and what I've done is I can find my right hand and my left left hand and I can box my face and I've also put a dot on my wrist so I found my right and left hands I boxed my face I boxed my face and I put a dot on my wrist so that's getting uh that's getting pretty impressive but what is the problem if you look up in the corner we are running at two frames per second or three frames per second and what that means is is that our programming skills and our capability have outgrown the processing power of the res Raspberry Pi so what we need is what we're going to need is we are going to need more processing power we are going to need that GPU if we're going to continue to do this type of work okay look it's starting to starting to get light out there in the old headwaters of the Nile River and it's still a little bit dark so the camera is in night mode but we should see that come back come back to life here pretty soon I mean come come to color pretty soon okay guys now what we are going to do and I'll stop in a minute for questions but let me just kind of go ahead and get into this let me switch over and we are going to now switch to the boot screen to the desktop of the most excellent Nvidia Jetson Oren and wow look at that is that not the coolest desktop ever that is really just I mean you you boot the thing up and you get really really excited okay now like I said say uh oh let's see Colin says I put media Pap on the Jetson Nano on 4.6 and it broke my Cuda yeah yeah I've played around with media pipe on the Jetson Nano and media pop actually is pretty fun and impressive on the Raspberry Pi but I it doesn't play real real well on the Jetson Nano at least with my skill set but never fear the thing much of what you can do with media pipe I'm going to show you how to do it on the Jetson products and be doing it down in the GPU okay so in fact we'll get into a little of that this morning or for you this evening it's early Saturday morning here for me so we'll get into that a little bit cold but yeah I've I've not had a lot of luck with media pipe and again I'm not saying it can't be done it's just with my skill level I've not been able to do it okay so again it's kind of like we've been given a Ferrari here and the Immortals can go around the race track in sixth gear at 200 miles an hour but what can us normal people do well the first thing rather than show you that I can do 40 trillion operations per second can I just get the thing booted yeah I got the card flash I got it started I got it booted and it actually went pretty smoothly I'm running jetpack 5.1.1 that's the latest jet pack and the thing kind of installed and booted up without a whole lot of trouble but now the first thing I want to ask is can I just access the cameras can I get it to like grab a frame and show a frame can I work with the different cameras that I have and you can see that I've got a lot of cameras here let's look at this right here in front of you I got one two three four five six and then in the studio seven eight nine ten eleven twelve so I've got a dozen cameras around here plus five IP cameras so all together today in this demo we've got 17 different cameras to work with and so we're going to make our way around those uh we're we're going to make our way around those cameras and see just what we can do and can't do as sort of medium skill mortal programmers okay so let's come back to our view of the Judson Oren Nano and the first thing I'm going to do is I'm just going to see can I run a webcam can I run a webcam will this thing allow me to run a webcam so I'm going to come over here and the first thing I need to do is just look and see where what cameras I have and what cameras this thing can see and so what you can see here is it looks like it's found three cameras this VI output I believe that's the Raspberry Pi camera that is connected to it that's that CSI connection and then you can see here I have a USB HDMI what that is that's a high-end Sony studio camera and then I take the HDMI out of that I have an HDMI capture board and then I feed that into the USB port of the Jetson Oren Nano okay and so that is not like that's not just trivial right there so I'm trying to see can I do that and then finally down here I've got a C270 HD webcam and so I've got kind of a standard webcam a really fancy Studio camera and then I've got the uh I've got the uh CSI camera the Raspberry Pi camera so let's see now that I see where those things are and what I want to see is I want you to see where am I okay let's see my mouse probably needs new batteries okay let's see if we can get this thing back together minor technical hiccup here but hopefully we'll get this silly Mouse to work again there it is back to live okay hopefully that battery will hold up but you can see that what I want to make note of is the webcam is video three so we're going to come over here to uh we're gonna come over here to visual studio code and what I did was I just wrote a simple program and this is based on the Jets and utilities I think uh Dusty Franklin the amazing The Immortal Dusty Franklin wrote the libraries and the utilities and so for us new guys it's pretty easy to get started by using his utilities so you can see I've got a lot of kind of nonsense in here of stuff commented out but maybe 10 to 15 lines of code we are going to see if we can run this thing now I'm going to activate this I'm going to activate this video three and that should be the webcam if I am thinking about this right so we're going to run this and just see if we can grab frames and show frames and see what kind of performance we're getting and so we're going to come in here and run it hold your breath ah boom look at that okay buffering here so let's let this uh let's let this catch back up okay it looks like it's coming back online sorry about the buffering guys I know it's uh annoying but let's try to power on through this I think I've got a good enough stream to uh to keep going here you can see that it looks like that we're about 1280 by 720 and we are operating at 30 frames per second and so one question that I have here one question that I have it looks like it's behaving pretty well one question that I have though is is that uh why are we only getting only 30 frames per second do I not have this set up right or is it a camera capability so I'm going to come up here and I'm going to look at that camera and that's video three I'm going to look at it let's see if I can move this where you can see it a little better and if I look at the capability of the camera you can see that on all the different orientations and all the different configurations of the camera the camera is only capable of providing 30 frames per second so there's nothing that I could do to get more out of that so the the the uh the Jetson Oren Nano is ready to rock and roll but the camera is just able to provide 30 frames a second so let's take a look at that again okay and there it goes now what I want to know is are are we operating in the CPU or are we operating down in the GPU and so what I can do is I can come over here and there's a cool little application called jtop and if you look right here jtop is showing me what the GPU is doing so it looks like the GPU is running at about 25 percent but if you look up here at the six CPUs they're operating at 10 20 15 something like that so I'm putting almost no load on the CPU the work is being pushed down to the GPU and that's what we want to that's what we want to see well if I go and just look just at the GPU you can see here that it is operating at about 30 percent so your GPU is just sitting there casually like saying is that all you got is that all you got give me more frames give me something else to do so just sort of an idle the GPU is able to do this and so that is pretty cool I think we're sort of where we would like to be on this I think we're sort of where we would like to be on this looks like we got some Visual Studio code fans uh Hey Dr Dave welcome okay so let's uh I'm kind of surprised my webcam is such a Ricky dink thing most webcams these days have uh more uh you know more resolution more power than that but I've got a bunch of different webcams and we'll be looking at different things now let's also go back and what did we say over here uh if we come back over here and we look at our devices again we look at our devices again you can see that that studio camera that studio camera is video one and so let's go see if we can put that studio camera on and so what I would need to do right here I would go from video three to video one and then I run that okay boom look at that now that is really high resolution it's coming in at 1920 by 1080 but then to display it on the screen where you can see it after it's brought in I downsize it I downsize it to 1280x720 to show but this is full 1920x1080 and we're operating full out at 60 frames per second so that is pretty cool let's see are we still staying down there in the GPU so we'll look here look now we're at about 50 percent on the GPU and we come in and we look at the CPUs let me come back here CPUs again are just sitting there kind of idling and so this is really what we like to see we are operating with the GPU and so that is really good now let's come in now this is where it gets a little more tricky I'm going to go out and I'm going to try to grab I'm going to try to grab some some IP cameras and understand if there's a few glitches in this it doesn't have anything to do with the Jetson product it's just with all these cameras I have I have just a little Rinky Dink router from the East Africa internet service provider and then I've got these 12 cameras I'm live streaming out the live stream comes back to me in a monitor and so there's a whole lot going on and what happens is when the router gets bogged down then it's not providing the data from the it's not providing the data from the IP cameras which are very high resolution it's not providing that data quick enough and then the program will crash and so that doesn't have anything to do with how the program is written doesn't have anything to do with the Jetson product just has to do with my Rinky Dink router and so let's come in here and let's see if we can snag one of those IP cameras okay so I'm going to come up here and let's see what would be a good camera that is I think I want to try 40 four okay I'm going to try my camera 44 ip44 and let's see if we can fire that thing up and if it doesn't fire up I'll try three times and if it doesn't work I'll move on and you will know why I've moved on boom look at that oh man it's raining out there we've got some fishermen that are kind of caught out in the rain and you can see that they are struggling pretty well to keep their boat going but uh these guys know what they're doing and I am sure they will make it back to shore safely don't stand in the boat don't stand in the boat that's what my dad always told me whatever you do don't stand in the boat well it looks like he's getting his uh it looks like he is getting his uh fishnet in okay let's see again this thing will crash in a minute because of my low bandwidth but let's just see if I come in you see I can point this camera around let's see if we got anything that looks like we got what we got here what have we got here ah two guys with machine guns how do you like that two guys with machine guns and you see what's uh what just happened there was uh that it's not feeding it's again my router and so I'm getting some of that aliasing because of the router let's see what else we have going on here see if we got anything else going on hmm okay but you can see that I can look around it's pretty cool don't you think okay but I digress so let's get back so I can look at an IP camera and let me and the thing actually didn't crash and so that's pretty that's pretty good okay so let's uh let's look here and let's see if we can try a different camera uh oh you want to see you want to see the front of the compound I will come over here and let's see if I can show you the front of the compound a different IP address and I will have to snag that IP address that one looks like 24. so we'll put in 24 and let's see what we see here [Music] let's run it [Music] ah there it is okay that that is out the front of the house looks like we got some people working out there okay that is our what I'm growing there is I'm growing comfrey for our rabbits and then I'm also growing Alfalfa in this uh this area this is the comfrey this is the Alfalfa and then this area here and I told you that would happen uh let's try it again this area here [Music] this area here is where I'm planting my corn and hopefully the corn will go in today and so we've got some people working out there in the garden let's see if I can see anything else interesting to look at ah another guy with the machine gun how about that okay so let's see I'm trying to see if we see anybody else out here that's my elephant grass I use that to feed the rabbits okay I'm just trying to find interesting things to look at so that when we go in so that when we go in and we actually try to do detection we've got some things that we can look at and try to identify so let's kill that and then uh let's see do you guys want to see anacam do you want to see anacam let's see if we can look at anacam which one is that that looks like 23. I don't know this one is a harder one to snag because it is a wireless camera and that just puts an additional load on that's puts an additional load on the uh the router so let's see here yeah look at that that's anacam you can see Miss Anna is having her breakfast down there very nice so you can see I got a lot of cameras and I've shown you a lot of different types of cameras and I can hook up to all of them and I can view them on this just being a new guy it's fairly easy to get this stuff uh uh you know to get this stuff up and going now what I want to do is the final things I've shown you I can do a webcam I can do a studio my camera so I'm going to come in and I'm going to run this thing and let's see what happens boom there it is okay now what I want you to see is let's let's go in and let's see what's what's happening on the GPU and the CPU and so you can see the GPU again is running at about half its capacity and we're not putting any load we're really not putting any load here on the uh on the CPUs and so that's good okay that's good now let's talk about sort of two things that are not not so great with what I've shown so far really if you look at this code if you look at this code you can see what I'm doing is I'm using the Jetson utilities and I tell it to capture a frame I tell it to analyze the frame and then I tell it to show the frame so really or you know here I'm just capped pressuring and rendering I'm just doing two things capturing and rendering and I'm using the libraries and the utilities that Jetson provides what's the great uh that Nvidia provides what is the great thing about that it's really easy to get it up and get it started the not great so thing the not the thing that is not so great about that is is that I lose the ability to customize and do the things that I want and so let's come back and let's look at this uh this image here all right if you see it's very pale and it's very washed out and it's only at 30 frames per second and as I go through and look and look and look I'm not able to find a simple way to call the camera using the utilities where I set the frame rate like maybe there is a way to do it but I wasn't able to find it and also there's all these sweet things that you can do in g-streamer and you can tweak the Raspberry Pi camera but when I use use the capture utility from Nvidia I kind of lose all those hooks and I'm kind of left with whatever I got then so that's one thing I lose the ability to really control the camera good news is very easy to get up and running bad news is now what okay I want to do more I can't really very well control the camera now you can go into the utility which I think is a c program and you can find where it is that's doing this stuff and you can modify it and you can recompile it I've done that before but it's just it's just too tricky it's just kind of beyond what is really my skill set so then the question is what do we do the the other issue is okay I say grab a frame and I show the frame or in a minute I'll show you grab a frame analyze it and then show it with the annotations like you know the things it's found and so forth but now I just sit there and now what what if I want to do things I've kind of lost the ability to control troll things and so what I would really like to do is I would like to have more control over the image quality and over the different things that you would be doing and so this is kind of the approach that I've been taking and then what we're going to see is can I do this can I do this and still can I do this and still maintain uh you know maintain most of the work being done down on the gpus does that make sense let me take some questions here let me pause for a second and take some questions okay uh did you guys lose audio I hope not okay you there was a little bit of buffering there but you guys tell me uh ask me some questions or tell me if the stream is going okay for you guys foreign [Music] let's see not seeing a lot of comments okay just me I think had to refresh okay so Brian's got it working that's good audio is working you guys asked me a few questions about what I'm talking about okay people are saying it's going fine but let's get a few questions in here what do you mean when you say operation one line of code no it's more like a floating Point calculation or an integer calculation and so some people talk about floating Point operations that would say like 3.2 times 6.4 is a floating Point operation or 3 times 6 is an integer operation enter your operations go faster but most of the things that you do with uh with detection and most of the things you would be doing down in the GPU would be integer operation so it makes sense to talk about uh operations per second as integer so that's a good question I'm excited about the Oren yeah I am too okay let's see can the live stream be powered off the Nvidia not really really because on the Nvidia I run wirecast for my studio and there's not a Linux version of wirecast and so I would not be able to do that maybe OBS would run on it but then if I'm running the studio on Nvidia then how do I show what Nvidia can do and so that would be something that would be kind of hard to do let's get a couple more questions in Andrew says my little brother wanted me to tell you he wants to learn the Arduino well tell your little brother to get out his kit and go to dopttechboy.com and get started really you want to start on the Arduino or probably today I would start on the Raspberry Pi Pico W and then from that you would probably go to the Raspberry Pi and then from that you would go to the Nano I would probably not try to start on the Nano there's just so much stuff you need to know before you do that so the Nano is a brick let's see I'm not sure the the Jetson Oren is a is is a brick in the sense of just a a heavy thing there but uh let's see is uh is Oren the latest Nvidia product okay the Jetson Aura Nano is the latest product the Oren agx has been out for a while and so uh this this is the latest product and it seems like they're phasing out the Xavier NX which I love so much but I look at the Jetson Oren Nano is sort of filling that niche in their product line and I really really love the Oren I'll tell you I'll tell you right now the Jetson Oren Nano I really love it okay you're a wild man okay oh Dave is that Dave is that Dave hey Dave I think uh we might perhaps actually have Nvidia in the chat I better be on my best behavior here better beyond the best what does the word Oren mean I do not know what the word Oren means so the Nano is usually no no the Nano is great I've got the Nano booted up I could okay all right you tempted me so I'm going to come over here and I'm going to live see if I can go over there boom that's the Jetson Nano all all up and booted right there and then I press the button again and I go back to the Raspberry Pi which was still running I should have I should have killed that program okay ah I think I have to press Q this bugs me I left a program running in the background okay there it is and now I press it again and I should come back to the uh Jetson Oren okay uh the Nano will never be useless that's right man and you guys like I say if you're just if you're on limited budget start with the Nano and then you older guys who want to impress your friends down at the hand club or whatever you can get the uh you can get the orange if you've got the if you got the coin for it okay quick someone find out what Oren means okay Oren is NVIDIA Amper GPU based okay very good thank you uh thank you David uh Greg mandiola is there any benefit of the official development kit versus a prediction okay guys uh I might be out of line here but it's just like I want to stick with the with the uh Nvidia products because in the support forums those guys are great and sometimes when I get hung I can go to the forums and they always respond someone from Nvidia always responds and they always work you through to a solution and so love that when you go to Seed Studio and if you go to the Nvidia forum and say I'm having this or this or this message naturally that the correct and proper response in the response they give you is well that's the that's the Seed Studio board you've got to go over there all right and then I don't know the Seed Studio provides a say so I want to stay in the mainstream I want to stay in the main line so I avoid the developer kits around the Jetson modules that are made by Third parties now I could be completely wrong you could disagree with me but that's sort of where I stand is there any of okay uh Nvidia makes super cool AI they really do and there's some really great guys there that uh that will work with you and help you make help you make things work okay those were the questions let's see if we can come back to uh let's see if we can come back and do a little artificial intelligence no I'm sorry we're still we're still looking at the camera so I left you with the issue in my mind of what the Nvidia products I lose the ability as a programmer to then start going and doing different things because I don't have the skills to go in and edit the utilities and recompile them and all that sort of stuff so this is kind of what I am doing and you guys can probably more understand what I am doing here and you guys that are taking opencv and so what I do now is I just create an opencv video capture camera just like on the Raspberry Pi and then when I come down here I grab a frame and then I resize the frame I convert the frame to RGB because that's what that Cuda core wants that core wants RGB so I convert it to RGB and then I send that frame down I send that frame down into the GPU with the Cuda from numpy this allows me to grab the frame in the way I want and then push it down onto that core and then once it's down there then use the most elegant Nvidia utilities here we're using the net detect to then analyze the frame and then it sends me the analysis back but I saved the original frame so now in opencv I saved the original frame when I sent a copy down into the Cuda core I saved the original up here now I get the result back I can decorate the original frame with all of our friendly things like CV2 put put text and then show and draw a box and things like that if you want to so let's run this and let's see what happens what are we going to run it on I think I'm going to run it on the webcam so let's see what happens if I run it on the webcam comes up and I hope this runs always get there it is ah I'm running the wrong one got ahead myself okay let me quit this let me come down what I wanted was the CV2 video viewer got a little ahead of myself there gave away kind of what direction we're going here okay so look at that that looks like that is the uh that is the webcam and it is running at 30 frames per second so when I'm doing it my way I'm getting the same framework maxed out uh maxed out frame rate as we were getting with the Nvidia with the Nvidia uh grabbing okay and so let's come over here now and look and you can see that my GPU is just idling it's barely doing anything and maybe we've put a little bit more strain on the CPUs but you can see I haven't maxed out the CPUs and so I'm not limited I haven't limited myself by doing a little bit more on the CPU but still doing the heavy lifting on the GPU okay so that is that now uh well set I didn't need to do that okay so let's quit out of this so I'm able to look at the video camera I mean I'm able to look at the webcam Now using my Approach can I go in and can I look at the high resolution Sony studio cam that's going through the HDMI capture board so that's what I want to see now and what I see is Boom there it is and we'll give it a second here to see if we can get any more out of it than that okay it's coming up and I am now operating at 43 44 frames per second on a 1920 by 1080 Sony studio camera and so that is pretty impressive again doing it the old capture CV way and then we'll pass it down to the we'll pass it down to the uh to the uh Cuda course and so let's see here uh let's come up and look at our J top and now you can see that I've got a higher resolution at a higher frame rate and so I'm getting a little bit more pressure on that GPU it looks like we're operating at about 30 percent of the GPU but 1920 by 1080 at 44 frames per second that is pretty darn good so we're going to quit out of that okay so what can we do we can grab a the frame and show a frame with the Nvidia products or we can do it the old opencv way and then push the data down into the gpus so now let's go on and see if we can do a little bit of object detection I'm going to come to this one though this is going to be using the Jetson way of grabbing the frame and then doing the analysis and then showing the frame and let's run this and see what we can find here so I'm just going to run that okay and you see up in the corner it's saying that it's seeing a desktop computer and so this is uh this is like image identification it's seeing and telling you what it sees it sees a keyboard and it sees a desktop computer and so that is pretty good now that is that is really great but the thing is with this it doesn't show you where it is to do that we have to do object detection I'll show you how to do that in a minute but using uh using this you can see that again it's not a whole lot of code maybe 10 maybe 20 lines of code so that's that's pretty uh that's pretty uh pretty good let's see if we can capture the camera stream I don't know if my internet is going to hold up well enough to do that looks like I've got a little bit of a latency here kicking in I'm going to give it just a second to see if I can let this latency clear out and then we're going to try to see if we can look at the IP camera okay so looks like we're back strong let's see if we can look at the IP camera [Music] okay it sees a Lakeside Lakeshore that makes sense let's see if we can zoom in our on our fisherman the it looks like the uh it looks like the rain is clearing up a little bit I'm not getting a real clear view of them but let's see uh canoe okay you see once they came into view it's seeing it as a canoe again that aliasing that you're seeing it's because of the speed of my internet connection but let's see if I can catch back up with those guys let's see if we can zoom in before they leave out of view okay it's trying to see a canoe but those guys are moving uh moving pretty fast yeah you can see that it's uh it's identifying it as a canoe let's back off and see what else we've got going on here let's see if I can come over here this IP camera I have is pretty uh pretty incredible it's a amcrest camera let's look across across the rainy morning on the mighty River now and that rain is going to make it a little bit harder to do an identification but let's see what it sees here if I can kind of jump in on this Cottage it sees it as a barn and that kind of makes sense because it doesn't look like probably in the training it had not been trained on this particular style of house let's see if we can find something else over there huh another quaint little cottage let's see if I can zoom in on that ah look at that that's that's roof okay guys that is pretty cool that is pretty cool let's take a look at our processors here so we'll come over here and we'll see how j-top is doing okay you see now that I'm doing the detection and again this is with the Nvidia grab the Nvidia analyzing the Nvidia show we are getting up to about 50 percent on our GPU and if we come back to the CPUs they're sitting there in there still just idling okay so that is pretty uh that is pretty darn cool but again I couldn't go in and tweak the camera very much I couldn't do anything it just sort of run what they've done which is cool but again I want to do more than that but I'm going to go ahead I'm going to show you the next step and that is instead of just seeing kind of like the main object that it sees what if we try to identify object identify where we not we we bought we find it we box it and then we say what it is and let's see if we can do that and I will start with video three there and we'll see what happens this one should work because it's not using the IP camera okay look at that wow okay we're operating at 30 frames per second which is as fast as the which is as fast as the camera as many frames as the camera can provide so it's it's full speed on the camera and look it's finding a keyboard it it sees a laptop let's see if I put the mouse here it finds the mouse let's see if I put a bottle here yeah it finds the bottle but the bottle covered up the mouse so we've got the mouse and that's a lot going on there's the bottle the mouse the keyboard the person so you can see it is doing a lot let's take a look at our let's take a look at our at our GPU and see what's happening okay now it looks like we are really starting to put some pressure on the GPU the CPUs are still just idling that's what you want the CPUs idling in now it's still like about 40 percent so it's about like it was on the object identification and so the GPU is saying they're saying is that all you got is that all you got you know give me more give me more and so that is going uh really well there so let's kill that but again here I am using the Nvidia to grab the frame the Nvidia to analyze the frame and then the Nvidia to render the frame and so if I don't like these boxes on here and the way they've done it you know there's a few parameters I can tweak but mainly I would have to go in and I would have to edit those utilities that they provided which is really beyond my skill set so let's go back and see how I would do this not you know being more of a less skilled programmer I'm going to come up to the CV2 detect which is where I am creating a CV2 camera I am grabbing the frame I'm converting the frame to RGB I'm sending the frame I'm converting it from uh to Cuda from numpy so now I've got the Cuda frame and then I do the detections down there on the takuda image I saved the original image and then after I get the data back from the analysis of the frame after I get the data back from the analysis of the frame then I go in and I decorate the original image that I grabbed using the tools that I know in opencv and so let's see how that works so what I'm going to do here is I think what I'm going to try to do is I do believe that that is the uh that is the high resolution Sony camera that we're going to go to here okay let's see okay look at that so it sees a person it sees a laptop and then let's see if I come over here it sees a bottle it sees a person it sees a laptop you just keep doing things till you break it sometimes it doesn't see a mouse in your hand very well because it expects the mouse to be on the desk and so it it doesn't it doesn't really see the it doesn't really see that but I've got bottle person laptop and I'm operating at 42 frames per second so let's see what happens when I come in and with the program I wrote we're still operating about a 40 percent on the GPU and if we come in and look at the CPUs you can see that we're putting a little bit more of a load on the CPUs maybe 30 percent but still it's not really impacting how fast I can run I'm still running really really really fast and so I think that's cool let's see now let's see if we can go in and turn on let's see if we can go in and turn on the the Raspberry Pi camera okay and so let's come here and you see that when I run the Raspberry Pi camera my way from opencv I have all of these different parameters that I can set to get a much cleaner a much clearer image and so let's run that and let's see what the Raspberry Pi camera does if I'm thinking about this right foreign so it's upside down because of the way the camera is mounted so in opencv it is a quick fix I come up here and I turn the flip method to two and then when you look at this long string it is applying that flip method in this string okay and then by doing that it takes care of it in gstreamer just like that okay let's see what happens [Music] okay so it sees a person it sees a bottle now the cups it seems to have problems with cups this one's a little bit kind of uh this one is why oh it's down here okay uh okay it's the bottle the person and this is kind of for some reason it doesn't like to recognize this as a as a cut let's see my favorite thing scissors seems like it's hard to oh there it is scissors it sees scissors it sees me it sees bottle okay that is pretty cool and look at that we are operating at 40 frames per second also do you see how much clearer this image is than the one I was getting by just using the Nvidia grab because I can go in and tweak it I won't take the time to do it but there's just one parameter I could lighten up the image a little bit also you see I did noise reduction so you kind of remove that digital noise from it so I love it that I can tweak that and then if I come over here I am still not overburdening the CPUs and so I kind of like doing that so let's uh let's quit that let's see if we can go up here and let's see if we can look around the compound a little bit see if we got any any activity going I will come back over here and I am going to go to I am going to go to which one of these do I want to do well let's take this one and then let's see which camera let's try let's try camera 24. let's see if we can snag that that see if we can snag that one oh Andrew sees my watch yeah you notice my watch matches my shirt I'm a fashion sleeve what can I say no okay whoa so what happened remember how I flipped it well we'll just flip it back put it back the other way we're going to put it back to where it was see see in opencv it's so easy for me to do all these little tweaks uh uh in the program so that's why I kind of like doing things this way okay let's see let's see if I can back up here and get something interesting I hope we can find something interesting to look at what happened did it not get the stream stream okay let's see what happened I've got camera 24 which should be right and uh let me try one more time maybe just didn't grab the frame or maybe I have a mistake in here somewhere I won't spend a lot of time debugging if it's not working we try three times on the IP camera if it doesn't work we'll move to a different camera there it is Boom okay now it thinks my uh it thinks my almond tree is a person but let's see I think I see a person there let's see if we can zoom in and see what that person is doing still working in the garden okay and it's having a little problem it's not it's seeing a person but it's not absolutely perfect because uh I think oh it looks it thinks the almond tree is a potted plant and then it sees it as a person but you saw for a second it said elephant and why would that be well she's got that hoe and now red looks like a fire hydrant okay so you see I would need to adjust the parameters a little bit which I could I just won't take the time I could tweak the parameters to do a little bit better but you could see that that hoe might make things look like a the trunk of an elephant or something like that but I would say that's doing pretty good we are doing the detection at 20 at 26 frames per second part of that might be that the stream is just feeding the uh the stream is feeding the uh the stream is feeding the frames not real reliably and that could be the lower frame rate on this because you see the GPU is not maxed out let's see how the CPUs are doing the CPUs or not doing are not doing that much and so that uh that is pretty darn good let's look around see if we can find anything else okay let's see this is going to be a hard one because this is kind of our guest house over here and let's see if we can see that chair okay just can't quite pick out that chair I don't quite have enough uh Russell yeah it's almost seeing it it is almost seeing that chair almost seeing it okay let's see what else we got here all right let's go around and see if I can find a couple of things here [Music] see if there's any ah I know what I can see let's see now this is just a partial of my pickup and let's see yeah look it sees a sees a truck okay look at that 30 frames a second recognizes the truck that is pretty cool let's see what else we could do here I just kind of showing you again I've just had this for a few days playing with it okay so let's uh let's put out of this and let me switch cameras we'll come over here and uh come down here and let's see if we can look at camera number 44 am I boring you guys or is this just kind of cool to see all the things you can do with this board and again uh the the thing that I'm doing here is I'm showing you what people with our skill level can do with this board and so the things that I'm showing you here I could show you how to do that okay we got somebody down there seems like this morning we're just seeing a lot of guys with guns let's see yeah recognize it as a person that's pretty cool let's see if we got anything else to look at banana trees let me see let me go to let me go to orange Cam you want to see orange Cam okay and what does it see it looked like it was trying to see an orange there let me see if I can help it a little bit got a little bit of aliasing going on in the Stream let's give it a second to stabilize and see if it can recognize the oranges on the tree sees a vase uh it is so close to recognizing that orange it is so close to recognizing that Orange let's give it a second you can see that it is almost seeing it okay there it is Orange Orange let's go to orange Cam 2. and let's see what happens okay it almost saw it let's move it over a little bit give it a little bit of help thinks it's broccoli that doesn't look like broccoli to me but uh but it is you can see that it's not orange and it's sort of with a lot of other stuff coming up so I could I could understand how that would I could understand how that would happen so let's quit out of that and then finally what I want to show you is I've showed you object I've just showed you crab Shore frame I've shown you object detection I've shown you object identification I've shown you object detection and now what I want to show you is I want to show you pose and this is kind of the type of stuff that media pipe will do and so as always I'll just start doing it with the Nvidia libraries and just see if the thing is working and we'll have to find a person for this to for this to work on but let's see how it's gonna it's gonna fire up the CSI camera so let's see what happens give it a second so this will be on the Raspberry Pi camera okay look at that all right you see finds my hands it finds my fingers the points on my face if I stand up you see it finds my it finds my various body parts and we're operating now it says 150 frames per second and that's just saying that it's probably going so fast it's reprocessing frames because the camera can't go that fast but that is pretty neat now again what do we not like if you I don't have the ability to go in I don't have the ability to go in and tweak the parameters very easily so let's quit out of that and let's see let's go to my program using opencv okay so let's go to CV2 pose and see what happens okay so you see again it's great that you can just call that routine but what if you wanted to label things differently or do things differently then the way I'm showing you is a little bit gives you a little bit more and then draw sticks around my face and then what I can do is I can show my hands where am I showing what oh I'm up here on the I am up here on the uh on the studio camera okay I've got my nose and I've labeled every single digit of my hand you know the my wrist my palm my thumbs every digit of my hand has been labeled and that is operating at 33 frames per second it's going faster let's see if it'll speed up a little bit now once I start you see top here I think you guys are getting some buffering so let me give it a second here I'm giving it a second I think you guys are getting some buffering okay I think we're back on not and now let's come and see if we can go back to let's see if we can go back to an IP camera let's go back to an IP camera see if I can find anyone okay so let's run the IP camera hopefully [Music] [Music] yeah David first time to see two machine guns we you know I have learned that it doesn't detect machine guns I tried let's see let's see if we can get something down here with this guy okay what is wrong with this it's not liking uh I'm trying to set this to camera 44 and I'm on CV2 pose and uh it says cam is not defined ah I let's see I need this on okay I commented out the wrong thing so let's come back and try again and I know you guys are getting buffering but I hope you're catching enough of this to be interesting [Music] takes a little while to fire this thing up sometimes uh and I got I've got to do the flip thing again so let me quit out of that and we're going to go back to flip equals zero and fix that so you see how easy it is to fix these things okay so let's see what I can do here now it's raining and this guy's a long way off but let's see if we can see okay there he is okay do you see how it's doing the pose on him and we're getting a body pose and that is pretty cool and I don't know if it's going to show his nose but he'd have to turn around and maybe we could even see his nose turn around turn around please turn around but somebody asked this guy to turn around let's see we'll give it a second ah got his nose Okay God is nose okay that's what we were looking for so let's see let's back off see if there's anybody else I think with the rain the people are Maybe on the porches somewhere let's see let me look at this other camera see if we can find anything so guys like I say I'm not doing the typical review here I'm just showing you the cool things that you can do with our skill level now this one's going to be a challenge of the lady in the garden ah we got we have we have a new player let me let me switch over to that camera okay let me switch over to that camera that would be camera 24. that would be camera 24. so let's see if we can switch over there and we'll come in and run it now this I don't think it's going to recognize this guy because he's in a really strange pose he's weeding the garden okay so he'd have to stand up for it to do the pose there it's trying to get it got his nose pretty darn good but if he's working there he's kind of he's kind of leaning over so we're not getting a real good example of it there it is it's trying it's trying a little bit but that's just not a very good pose so let's see in this other Garden standing behind the tree so it's kind of hard because she's standing behind the tree we'll give that a second and see yeah you see it's starting to get her and it does seem to be finding uh hands where they're not but there it walked uh it walked out let's see if we can follow her yeah call that a success what do you think okay guys I've shown you a lot of stuff here let's take a few minutes and let's ask some questions I've kind of I've been really I've been really busy here with the uh I've been really busy here with the uh uh and I want to see I I'd really hope there would have been more action on the river but you know it's it's very dangerous on the water when it's raining ah we've got a let's see if we can do that okay hold on hold on I'm trying I'll take questions in just a second you guys be thinking of your questions and let's see if I can do this quick enough we got a little action on the river and let's see if we can capture that uh not quite he's you see he's behind the uh he's behind the uh and he's sitting down so it's having trouble it's having trouble recognizing those as people because they're sitting down in a canoe with uh all right let's try okay let's try this I'm going to go back to CV2 detect and then I'm going to go to camera 44. so let's see if we can see some action on the river oh you look person person and boat okay that is pretty darn good I would say you can see he's bringing his Nets out of the river let's see if I can come in a Little Closer okay person person vote boom okay and again this is this is the things I'm showing you or with the skills that we have man are those avocados look at that avocado cam that tree is loaded now I know it doesn't recognize an avocado but look at that I'm gonna be eating guacamole tonight that's for sure okay let's look and see if we got any other generally useful or interesting things going on on the river something we might be able to identify like I say the fishermen do not like coming out the fishermen do not like coming out in the uh in the rain because the river is quite dangerous at that time let's see if we got anything down here okay it's kind of quiet I think uh I think people are up see if we can just recognize this as a person this guy's got a pretty mean looking stick I don't know what happened to the machine gun that kind he's pretty serious security guy isn't he what do you think yeah okay let's see if there's anything else we can find just to do some detection bench look at that found the bench and then I'll come over here usually we got some dogs running around [Music] am I just boring you guys with this I just think this is really cool stuff here chair chair all right so we got bench we got chair we got uh we got uh we got people and so that my friend is pretty cool stuff let's go back to the River View and I'll come out to the uh this View and if we see anything if you see anything interesting we'll try to zoom in on it the guys with the machine guns are in the front but I won't switch back over there if you guys see anything interesting let me know and we'll try to zoom in okay I've done lots of talking and I've not interacted with you guys as much as I have so what do you think is this interesting stuff for you guys are not interested in it what do you think let me look at some of the some of the questions uh what camera is it using the camera outside is an amcrest panzoom tilt IP camera and I think it's about 350 bucks and man it is a very very capable camera for the price and I've just had a lot of fun playing with that all right what processes are you guys using uh the Jets and automate I'm not sure I understand that can you do a course on this that would be very amazing that's what I need to know from you guys I did that original Jetson Nano class but that was like back on Jetpack 2 or something like that and a lot of things have changed a lot of developments plus I know a lot more now and could do a lot better job and so if you guys are interested let me know like I say your entry I've shown you the the deluxe here I've shown you the the about 500 uh Jetson Oren which what's my review two thumbs up I got it out of the box I spent a week playing with it and this is what I've been able to figure out in a week I figured this out in a week and so because of that I feel pretty confident that if I spent more time on it we could do some really cool things of course like one thing I did I don't show it because I couldn't get it working perfectly but I actually tapped into that amcrest camera from opencv where I could move the camera and so if something was walking across as something was walking across what I could do is the boat was going by I could Servo the pan tilt camera on the boat and I could track it from opencv and that was really cool but the problem that I was having is is that it it that as the camera moves you see how I'm still getting some aliasing in the image and as the camera moved is when the a-listing came happens and so the water's moving and then the camera is moving and you're getting aliasing so it was kind of hard to get good object detection it was kind of hard to get good object detection when you are when when you are you know having a not perfect signal coming in so we've got a boat and we've got person that's pretty pretty good stuff there let's see I got somebody else let's see they're a little bit behind my elephant grass and so let's see if we can snag those guys it sees the people but it doesn't see the boat but that's kind of because the boat is being covered up a little bit let's see what else we can do here okay so I'm sorry I got I'm easily distracted it thinks that that house is a kite which I'm not sure why okay so what I need to know from you guys is are you interested in this this board that I'm showing you today which is just spectacular it's the you know it's the formula one but the Jets and Nano I could do anything that I've shown you here on the Jets and Nano I just couldn't do like five simultaneous cameras or something like that if you guys are interested in that and me redoing a class on this let me know okay how much is the or in the orange it's about 500 bucks I don't think that you can order it yet is as of the last few days it was kind of like you could pre-order uh but you're starting to be able to get the Jets and Nanos on Amazon for 150 and so if you're young and struggling the Jetson Nano is really great if you're an old coot with some coins laying around the uh the I would definitely recommend the Jetson or a nano okay let's see what we've got here uh Devin says it's interesting is it buffering for anyone else yeah you guys it's just when it rains I don't have a good internet connection and so you guys are getting buffering because I don't have the best internet connection today uh Dev uh Dev start with the yeah definitely start with the Arduino lessons you don't want to start with the Jetson Nano okay uh programming isn't as hard as it seems okay thanks Andrew uh who are you I'm I am uh okay let's see great stuff Vaughn says this is fantastic keep going okay guys appreciate the feedback what camera answered that how much is the Oren I answered that please do another Judson class okay Colin wants to see another Jetson class Vaughn wants to see Von you already have the Jetson Nano right so if you're interested in me doing a class do let me know BBF says we could do amazing things yeah I think you could especially I haven't talked about it but it's those gpio pins it's like I'm doing all of this stuff and then I've got gpio pins and then you can control servos and do all types of stuff so it's pretty exciting I'm digging the new okay uh oh slit back here yeah well I just got up so I put a little little gel to keep the hair from being quite as crazy as it would have been if I hadn't I've glued my hair back a little bit okay let's see what else we've got uh why is it not identifying anything uh not a it just doesn't it doesn't seem to really see trees it just doesn't like there's a whole bunch of different types of trees and it just doesn't see them now you can see a boat back there but I'm not zoomed in enough you have to have a certain size image before you can see so it's got to be a certain fraction of the frame before it'll see it but there it's seeing the people and let's see if we give it a little more help there vote person person okay so it's seeing that okay you guys ask me some questions uh uh BBF says check out uh deep blue uh almond wants to get a Jetson but I do not understand their lineup okay guys just simple uh almond start with the Jetson Nano okay just get the Jets and Nano 150 bucks everything that I am showing here I could do it the Jetson Nano but like with the Oren I could be running like five IP cameras at the same time and doing different things on them but the Jets and Nano certainly is going to be running one or two cameras if that makes sense I I've not used the Teensy boards uh uh Vaughn wants to see another class would be great I think I would order the Oren okay Von's one of these guys that you know if there's a new board to order he's going to order it and so he was probably one of the first guys to get the Jetson Nano and went through my first class on that how many gpio does it have it the Jetson Nano has 128 cores and will do a half a trillion operations per second because you see in everything I did today I never maxed out the GPU and so if you were doing these things on the Jetson Nano you would be getting similar results because we never maxed out the Oren and the things that we uh did let's reboot the Jets and Nano series Curtis says uh oh how many gpio does it have it's I think it's the same as the Raspberry Pi I thought you were asking about the GPU course okay uh can you install matplotlib skippies I don't know you can put Jupiter notebook for sure and most anything you can that python would do what you could do okay on the uh on the the Nano the one thing like I say that I haven't had a lot of luck with is I haven't had a lot of luck with media pipe because media pack doesn't play well with opencv and when I installed media pipe it goofed up my opencv where the opencv then wouldn't run well with the Jets and stuff and so given that you can do the pose stuff that I showed you I don't think you really need media pop at least that's the uh the way uh the way I look at it up Andrew it's his bedtime Andrew is a student so he will have to get up and go to school tomorrow so he's going to go Andrew says goodbye so the Oren agx has like 2 000 GPU cores yes the or the Jetson agx has 2048 Cuda cores and 12 CPUs if you guys want to see a live stream on the agx even though you know it's probably not in the price range of what most of you guys would be interested in if you just want to see what that monster what that bad boy will do let me do know and we could do a we could do another uh we could do another live stream on that let's see the Jetson Nano is great guys you will learn tons unit using it yeah on your right quality over quantity can you run a program on two cameras at the same time you can have I I had like this thing on the ore and I had five cameras going and at the same time and each one was doing something different and it was just like five cameras is that all you got and so yeah that that was uh that was good uh let's see what are you doing for the rest of your day if you're asking me I uh I've got a new guy coming on today a guy from attesso the attesso tribe is going to come on and help us out on the compound and so I'll try to get him comfortable and you know plugged into the compound and he'll be helping us here okay good night Andrew see you guys okay any other questions guys we've gone a little long today because we had lots of stuff to show you but let me give you one more chance to ask some questions and then if not uh let me know uh you know let me know if you'd like to see the agx in action sort of like the stuff that I've shown you today if you wanted me to do the agx maybe we could see just how far we could push that thing just out of interesting is it better than the Raspberry Pi 4 oh yeah like I said say it's 100 times more processing power than the Raspberry Pi 4. can I use the Raspberry Pi camera module with this yes I've shown you we've been using you know like every third thing that I showed you was a Raspberry Pi camera on this so yes the Raspberry Pi version 2 will work on the Jetson or Nano you just have to have that little cable adapter that you would use it uses the same cable that you would use on a pi zero okay it uses the same cable you would use on a pi zero okay we got our fishing guys kind of have gone out of view let's see if I can find another thing you guys asked me another okay uh where do I go to learn uh Devin you want to start with the uh Paul mccorder Arduino lessons and after you've done the Arduino I would probably you I would probably then go to the Raspberry Pi unless you know a little python already and then once you kind of can do those basic things then you move to the Nano ah we've got action we've got a person and we got two boats does this guy have a machine gun this guy has a machine gun foreign doesn't see the gun okay let's look and see go back and see if we can find the boats I'm sorry I'm so easily distracted here we got some action let's zoom in on these guys we got a boat person person okay uh I'd love to see the agx okay maybe we'll do that let's check out the agx how do micro Mouse car how to do okay I don't know how to do that let's do something amazing computer vision yes that would be good okay so that is pretty good let's go and I can't help it guys I am just I'm probably boring you to death but I am just so fascinated with this stuff let's go to the front let's go to the front and what was our front cam our front cam was cam 24 and so let's see if we can come up ooh that's not good you guys ask questions while I'm doing this and I'll give you some questions and if I'm boring you I'm sorry it's just I think this stuff is so cool and again we watch Dawn at the source of the River Nile and we have looked at an amazing new board let's see if we can see this uh lady is no longer behind the Almond Tree let's see if oh yeah Caesar is a person uh we got another person I saw another person ah just out of view let's see if it'll snag her ah just can't quite ah got a little bit of a view of her okay let's see what else I can find interesting back here let's see if the guy in the garden I think the guy in the garden has sought shelter and it's not oh there he is but he's behind the Cucumbers okay I'll give it a chance to catch up he's behind the cucumbers he uh he's very shy guy he tends to hide behind the Cucumbers okay let's see is that live video yes Jack this is like do you think toptechboy.com is going to show you a canned program I'm sorry I should be on my better Behavior but I get tired really of seeing that same video Loop of 57 cars driving by on the highway and it identifies and counts the cars and all that that's great but that's something that's really tailored towards the road is sitting there the road's not moving it's looking at cars and the cars are driving by and there's a really good internet connection and I mean it's a it's from uh it's from a video file and so yeah that works great and you can run that demo program but then when you're doing things in in the wild here like this you start seeing more of the Practical problems okay we gotta it was seeing him as a person so that's good uh we'll see if it can uh kind of see as he gets up you can see that it's finding him as a finding him as a person okay uh let's see how does the graphic user interface work well ah look at that eggs person with eggs person with eggs although it thinks the eggs are a frisbee sort of disappointingly let's see if we can track the action here see if I had those servos working better it would just track her and it thinks she's carrying a frisbee uh it wasn't expect expecting to see a lovely Miss Rose an African woman carrying eggs that's probably not something it trained on okay so that was pretty cool we'll come back over here I'm just giving you time to ask uh ask questions here as we go on so let's see uh I wonder it's one of your most amazing things I've ever seen okay BBF appreciate the feedback there I wonder how powerful the agx is compared to the human brain ah there goes someone does he have a gun that's always what you want to ask does he have a gun no okay let's see if it'll catch him both yeah two persons that's good I'm gonna try one more time on that chair I just get get obsessed with these things yesterday when it wasn't raining it was finding that chair but today it's not going to find the chair okay let's see uh what else can it do yeah it can do a motion recognition that's like no problem at all did you get the pi 0w with camera working on the Jetson Nano yes I did so I have the pi 0w with one of those cheap little cameras and then it rstp feeds the data back to the Jetson and those little cameras and Pi zeros are so cheap you could put them everywhere and then you could just be getting gobs of data into the Jetson Nano from all over and that would be that would be an interesting uh interesting project top Tech boy for the win okay thank you BBF uh let's see uh hi Paul please teach us to program robotic car yeah the problem with the robotic cars they just don't work do we have a new player here do we have a new character no he's the he's that's curia he was working in the garden and now he's going on a little walk okay uh yeah the robots that we can afford just mechanically they're not stable enough so like you've had the robot stuff but I just end up getting we get uh we get loads of of uh we get loads of eggs and the cool thing is we get all the eggs we want and then the people that you see working out here we share eggs with them and on people that are just operating at a sustenance level to be given a dozen eggs every few days that's a game changer that's a game changer so we try to be very kind to the people around here how many IP cameras are you running and what brand are they I'm using amcrest cameras and I think I have six IP cameras around the compound now I could run like I was trying a couple of days ago to run them all at the same time and do this detection and the Oren has the the UMP to do that but my network my router doesn't have the UMP to manage all that data at the same time can it detect a drone or a big bird it does detect Birds I don't know about planes okay what institute do I need to go to in order to understand this you need to go to the Institute of top techboy.com just go to my channel man I show you how to do all this stuff everything I'm doing here I show you how to do uh let's see how much will it cost won't cost you anything at toptechboy.com this channel is right here yeah free okay guys it seems like the questions are slowing up this has been a very long live stream and so this is probably I'll give you one last chance to ask questions to see me go right we've got the Raspberry Pi lesson class going long-range data transmission and those little you know five buck IP cameras so it's like for 10 bucks you can have you can make your own IP camera with a uh with an esp32 board something like that and then you can do long range data with the Laura there's some exciting stuff out there and I would be very interested in this I would be very interested in this is NVIDIA paying you no Nvidia did give me this board Nvidia gave me the board but then you know they shipped it to a freight forwarder for me in the United States and then I had to get it from there over here so they gave me the board but I ended up having to pay like two or three hundred dollars to get it over here to my desktop so I paid more than Nvidia paid for the board to give it to me for free but they do not pay me thanks for teaching Paul everything is for free esp32 Cam is okay Oren means pine tree okay I did not know that okay guys let me let me say one more thing before we wrap up here how do we go to the next level here so with opencv the reason I love it is I can grab the frame push it down to the Cuda course do the processing send the data only back up and then decorate my original image or annotate my original image in opencv and then in opencv I could interact with the gpio pins to do cool things with servos or Motors or all that sort of stuff so it's kind of like all types of things can happen but how do we take things to the next level the next level would be deep stream and I tried for years to understand deep stream and I could just never get my head around it but actually a few weeks ago I actually got some stuff working in deep stream but like the way we think is normal human Mortals we think of writing a program like go grab a frame do something with the frame show the frame like do this step do this step do this step and the thing about deep stream is it's kind of like a totally different mindset there is this application called Deep stream and it's not so much you write a program as you have a template and you fill the template out and that template is what camera sources do you want to do and then the next part is what analysis do you want to do and then the next part is where do you want to send it and how do you want to show it so you're just filling out a little text template and then deep stream runs on your text template so you've got to think about things completely differently and then that is just blindingly fast because everything is done down in the GPU and it's just mind-boggling what you can do in deep stream and I actually I got it working and I got it working with one of my uh with one of my cameras and so that would be kind of the way for us mere mortals to take the next steps is to kind of finish out what we're doing here in opencv and then really start doing deep string and if there's enough interest I could try to learn dstream well enough to actually teach you guys and the things that I saw that I was able to do with deep stream is really amazing so let me give you an example of what the way you would think about deep stream like if I'm writing a program and a house is on fire how would I write a program walk to the fire hydrant open the fire hydrant put water in it close the fire hydrant walk to the house dump the water on the house go back to the fire hydrant so you see that's writing a program but you see with opencv I mean with uh deep stream it's more like this you're saying connect a fire hose to the fire hydrant point the fire hose at the house and then just turn the fire hose on and so it's sitting there and it's just doing the thing based on the parameters that you set up and doing things that way it's just absolutely amazingly blindly fast okay let's see what's deep stream a neural network no deep stream is it the best analogy I could give you to deep stream is G streamer we're in gstreamer I can create like a like I can it's like you're taking blocks and you're plugging them together and then your data goes through and that's kind of what deep stream is you're putting these little blocks together and you're putting the parameters for those blocks and then deep stream just runs a little bit like G streamer I wish there was a better way for me to answer that has Nvidia made any easier to do Jetpack upgrade I don't try to upgrade jetpack what I do is I just flash a new card I save my programs and then I flash a new card and then I do it that way this over the air stuff it was kind of like something I always ended up kind of getting messed up and that's not saying there's a problem with it that's just saying I'm not talented enough to be confident that I can smoothly do a Jetpack upgrade over uh you know on the board itself oh don't have the nodes or computer processing for this I have a little phone no personal computer how would I analyze through AI without the equipment well at some point you gotta kind of have the equipment okay uh let's see uh Devin try the app solo learn hey does your camera work at night time also some of the cameras have I mean they're all IR cameras but if you're looking at some of those things like if you look at this image here the IR illuminators would not be strong enough the IR illuminators would not be strong enough to you know uh to illuminate the whole compound you'd have to think of the IR Aluminator as a invisible flashlight and so if something's close to the camera it will work okay uh let's see uh Devin I guess I spoke a little bit you can you could do deep learning if you got like an online Linux machine that was like a powerful one and then you could configure it to do this sort of stuff and it's like penny per minute or pennies per hour to use it and so I guess that would be an option but I am not ready to teach that can you buy the equipment online I'm not sure Colin which equipment uh why would I buy a camera when you already have one okay I'm not exactly sure what that means and then imagine what you could analyze with a heat vision camera yeah I had one of those little heat vision cameras but the thing is they're so expensive the one I got was about 250 dollars and it had like 60 pixels and so when you're talking about high resolution if you were talking about high resolution infrared cameras you're talking about many thousands if not tens of thousands of dollars if that makes sense okay guys I swear I was gonna I was gonna wrap up leave comments down below if you want more live streams I also bought the Raspberry Pi high quality camera in like six lenses for if you guys would like to me put see me put that through the pace let me know if you'd like to see me do the orange agx Lem just kind of to see what that thing can do then uh let me know and uh we can try to get another we can try to get another live stream setup okay guys this has been a lot of fun hope I didn't bore you hopefully you find this stuff interesting but I'm going to close it at this point appreciate if you guys would leave a comment down below give me a thumbs up and share this with other people because the world needs more people doing Technology and Engineering and fewer people sitting around watching silly cat videos Paul mccorter with toptechfoy.com I will talk to you guys later [Music] [Music] foreign [Music] um

Channel: Paul McWhorter

Views: 5,713

Keywords: STEM, LiveStream, TopTechBoy

Id: 21785iuTkyU

Length: 96min 38sec (5798 seconds)

Published: Sat Apr 01 2023