AI on the Jetson Nano LESSON 49: Installing NVIDIA Object Detection and Inference tools

[Music] hello guys this is polemic order with top tech boy calm and we are here today with lesson number 49 in our incredible tutorial series where you are learning artificial intelligence on the Jetson nano I'm going to need you to pour yourself a nice big mug of iced coffee and I am going to need you to get out your Jetson nano gear and I'm gonna need you to get ready to learn some cool new stuff hey as always I want to give a shout out to you guys who are helping me out over at patreon your help is what motivates me to keep this great educational material coming you guys that are not helping me out yet think about looking in the description down below there is a link over to my patreon account think about hopping on over there and hooking a brother up but enough of this shameless self-promotion let's jump in and talk about what we are gonna learn today but first I must give you a little heads up you see this haircut yep the barber shops are closed still as of the time of making this video due to the coronavirus and therefore I finally got so sick of shaggy hair that yes I got out the old dog clippers and I cut my own hair with dog clippers so you might not like it but at least it's not shaggy leave a comment down below how many of you guys have also had to either cut your own hair or find a family member who was brave enough to try to cut your hair not perfect but it's better than getting better than having shaggy hair but enough of this small talk let's jump in and talk about what we're gonna do today okay so so far in the first 48 lessons we've done some really cool stuff and we've gotten up to the point that you can not only find faces in a picture but you can train it to know who's who and you can not only detect a face but you can recognize a face and that's pretty neat stuff but so far all of our work has really been based on open-source types of resources libraries like open CV running inside of Python those libraries we were using for face detection and face recognition but those are all sort of generic platform type of resources but if we really want to take our artificial intelligence to the next level we need to start using some libraries that were developed by Nvidia specifically for the Jetson nano because for example there's a lot of things that you might want to detect besides just face detection and face identify identification you might want to find other things and you might actually want to begin to train Network so to start taking our artificial intelligence of those next levels it is time for us to start moving towards some of those Nvidia jets and specific packages and so that's what we're going to be doing today is we're going to be installing some of those Nvidia libraries so it's going to be it's going to be one of those lessons where we are doing a certain amount of installations and you can follow along with me step by step but I will try to at the end go ahead and write one program and the simple program that I will be doing today we'll be looking at object identification where we're going to be putting some some objects in front of the camera and see if we can identify what it is so I will go ahead and I will get out of your way and then I will be firing up a terminal and that looks like it is already pretty much in the right position so we are going to begin by doing a sudo apt-get update to make sure that we have the latest and greatest of everything how long this takes is going to depend on how long it's been since you've done an update if it's been a long time this could take awhile if you've done one recently as I have then it doesn't take very long so it looks like we are ready for the next step and what we're going to need to do is we're going to need to install some some software that will allow us to do to install these libraries so there's some tools that we need so we will need to do a sudo apt - get sudo apt-get and then install so we're gonna do a sudo apt-get and install like that and what we need there's a program called git GI T that allows us to get things from github we are going to need C makes so that we can compile things we're gonna need a Python library that is called l ib py th o n3 - de Vie and then python 3 - numpy ok and so take a quick look at that oh my goodness I know you guys hate it it nothing makes my audience more unhappy than for me to be sitting in front of what I'm typing so I'll let you take a look at that Python 3 - numpy those are the things that we are going to install and so we're going to go ahead and click enter and these really usually don't take very long to install so I'll just sit here with you whoa lied Python 3 oh I put two dashes my bad hopefully you caught that library live python 3 1 - de VI we'll take that out and then these hopefully won't take very long to install it looks like I had already installed these so yours is going to take a little bit longer but probably not very much longer than what it took me okay this I can't remember if I mentioned but this what I'm doing is I'm following a tutorial that was put together by dusty at Nvidia and want to give him a shout-out because he really took some very very complicated software and reduced it down to some libraries that can be very easily called from Python so he took some very powerful packages and reduced it down to some simple libraries that were able to use very powerfully so shout out to dusty okay so we did our we did our installations and now I'm going to go ahead and CD down into downloads okay CD into downloads I guess it depends on where you're starting so I'm going to do a CD home like go home with the squiggly and then slash downloads and that will work for you no matter where you are okay now if I do an LS I am down into downloads and you can see some of those libraries that we had installed before it looks like the Kota OSS you remember we did that and then the D Lib was one of those libraries we needed for face recognition so we've been it down in this folder before but we're gonna go ahead now and put in the resources to to tense all these these in video libraries so what I'm gonna need you to do is do a GI t get GI t get and then clone we're gonna clone and then we're gonna clone - - recursive now what does a recursive clone mean that's when we go to github we don't just get the main folder we dig down into the folders into the folders and we get all the folders and all the files that are below the roots so recursive means you're going to get everything that's there is part of this package and then we're gonna do go to http the location of all this stuff / /g ith ub github.com / dusty - envy do you see that dusty - env were almost there / Jetson inference okay and you could see all that Jetson - inference okay get jetty clone space - - recursive space HTTP colon slash slash github.com / dusty desh envy / Jetson - in fronts with a little look boom it is cloning I think this is gonna work I'm just gonna wait cuz this just takes a couple minutes and so we'll just sit and chat while we're waiting for this to download and I think I can now safely go back to this view without having you yell at me it's going along pretty well here then after this what we'll do is we will start actually we're going to have to compile some of this stuff that we're downloading in so okay that looks good so we're still in download and now if I do an LS you can see that we now have a new folder which is called Jetson inference so that is where it put all of that good stuff so we are going to CD down into Jetson inference that we just created okay now that we're in that folder we need to make a build folder so we're gonna do an mkdir the UI LD and now LS and you can see that we just created this nice build folder okay and so we are going to CD down into build so we will be a building in the build folder and now we are in build that looks good so we're gonna do AC make and then we are going to do a dot dot and a slash and this should start putting these packages together with a little luck looks like it's going okay now what I do believe is I do believe that this could take like 30 minutes but in the middle of this it's gonna stop and it's going to come up with like a green and white dialog box and that dialog box is going to ask you for certain things to install and so what I'm gonna do is I'm gonna pause this operation and let all this stuff happen and I'll come back when that green dialog box pops up asking you for what to download okay and so I will check back with you in a minute okay actually happen quicker than what I thought it would but we will go ahead here and this is these are pre-trained models that allow you to do certain things and these models are very complicated I mean these models would take very very large amount of resources to put together and much more complicated than what we could do by ourselves from scratch so we're taking advantage of artificial intelligent models and training sets that other people have already put together and so the good news is they are free and so on these models I'm just going to come up to all and with the all selected if I hit my spacebar do you see how it puts an asterisk there that is going to get all of the image recognition models but there's another type of model so I don't need to come in and select each one of these because I got them all but there's also detection models so you've got to carefully come down and then come down and you see here are the detection models so I'm gonna get all detection models okay and then we are going to come in and then there is semantic segmentation I'm not going to select all of those I'm just going to take the default ones which they've already selected which is 24 25 and 27 and 29 and 31 and 33 well they got already a lot of those but I'm just gonna I'm just going to keep the ones that they selected I don't want the semantic segmentation legacy model so I'm not going to choose any of those and then finally the image processing all models I'm going to get that all of them which would be the deep homography cocoa and the super resolution okay and so those are interesting interesting models I will just take a second and say what is the difference between image recognition object detection and semantic segmentation okay image recognition what that is you try to have kind of a clear field of view and then you put something like a cup and so you would want nothing in the picture but the cup and then it would identify it like what's in the picture a cup or a pin or a bottle so you show it a kind of clear picture of something and it tells you what it is that would be image recognition now object detection is a lot more complicated that's like where you would look at the room and it would find in box all the things that recognizes like oh here's a bottle here's a chair here's a person and so object detection is looking at a complicated scene in finding and localizing or boxing a lot of different things to where the image recognition is a lot simpler where you need to kind of show it clearly a picture of something and then it tells you this is a dog this is a cat that type that type of thing okay so that makes sense so we are going to go ahead and make sure you come down with your arrow key and B over okay and then click enter now we are downloading a lot of stuff here so this is actually where I think it could take 45 minutes and so this is where I'm now going to pause the video and after all of these things get downloaded then I'll come back because the next thing that's going to pop up is it's going to ask you some questions about PI torch and other one of those ring dialog boxes I'll come back and help you with that in just a minute and we are back and as promised is asking us about PI torch and yes indeed while we are doing our installations we might as well go ahead and install PI torch what is pod torch pad torch is something that allows you to sort of retrain these networks that were downloading so if we wanted to create some custom image recognition capability we would start with one of the existing pre trained bottles and then we would retrain it and tweak it for our specific application we'll talk more about that later but we're going to go ahead and install PI torch you see the first option is PI torch 1.1 for Python 2.7 we want to not do that we want to arrow down you see how I'm arrowing up and down you want to arrow down to the PI torch one point one point zero for Python 3 point 6 and then spacebar puts an asterisk there we should be good to go if I click enter that is on the ok the OK is highlighted so with a click enter it should go and let's see what happens here it looks like this is going all right I will check back with you after by dortch's installed and it looks like that installed everything without trouble and so I do believe that we had just done the C make command and so now after C make we need to do make so we will do make and then I - J dollar sign and then the letters in P are o so we're gonna do a make - J dollar open parenthesis in P our o close parenthesis and let's see if this is going to be one that takes a long time ok it looks like it is going I can't remember I've done this a couple of times but I can't remember which are the long ones and which are the short ones this might be another long one and so I will let you watch your screen by yourself and I will catch up with you and and it looks like that worked without incident and we are getting really really close to getting this done we do know we've done the the sea making in the making we need to now do the installing so we will do sudo and then make and then install like that sudo make install you can see that completely enter and this is flying through this this might go quickly so let's see if I maybe we can just sit in chat as this one goes through but what we'll do here we're getting really close after we get done with this there's one more command and then there's another little helper software that we're going to go ahead and stall because sometimes these installations so let's just take a video and get it done okay that worked good and that was the sudo make install now we need to get everything synchronized so we will say sudo LD kind of fig and this should just run really fast sudo LD config yeah that went really fast and now there is another software that that with these reference things we need to kind of understand what the parameters of our webcams are now if you're using the Raspberry Pi camera you know that as soon as you launch that GStreamer command it gives you that printout down below of all the different frame rates and resolutions that the camera is capable of and so for the Raspberry Pi camera you can just grab it off of that when you launch it it gives you what your options and parameters are but if you're using a webcam you need a helpful program to help you find the properties of the webcam and that is the program v4 l dot utils and so I need you to do a pseudo oh you guys are going to get mad sudo apt-get install v4 l dash utils okay and so we'll hit an enter there and that is building and that got installed very quickly and now what that program is going to let us do later on it is going to let us sort of query our webcams to see what resolutions they're capable of and therefore we'll know suitable parameters to write into our programs okay I do think one more thing we need to make sure that we got the PI torch and or that yeah the pipe torch installed correctly and so let's go ahead and just hop in Python 3 and we'll get to the Python 3 interpreter and what we can do is just type in import torch and if torch was was if pet torch was installed correctly you see it does that without an error now if you got an import torch error that like torch wasn't found then your torch didn't install right and you've got to go in and kind of figure out what happened the other one we needed was torch vision which should have come with this torch vision ok so we've got things really good we've got a nice good jetpack 4.3 installed and we've got these inferencing and object detection and image recognition libraries installed and we have pi torch installed so those are all very good things plus we got that VL 4 program installed so i'm going to do a ctrl d to get out of that and that all looks really good you know what I think that we are ready to try to go ahead and in write a quick program ok because after all of this installation we do and we do kind of owe it to ourselves to have a little bit of fun so I do have the PI pro folder I'm opening up code OS us and then I have created a new folder let's see this new folder is called Nvidia ok and inside the Nvidia folder I am going to create a new file so I'll come up and I will put a plus for a new file and then I think this is going to be my and then I will call it my identify - first for my first try dot pi something like that that's kind of small for you guys to see but you can name a file on your own right and now we have a fresh new ready to go Python file that we are ready to write a program okay now one of the cool things that we just did was we installed some new libraries and we're going to be using those libraries so I'm going to import jetson da IND for intz jetson dot inference and and import jet sunt dot util so those are the two libraries we are going to be using just for fun you're always kind of worried did they really get installed so I'm going to right mouse click run python file in terminal and this doesn't do anything but it did load those libraries and so that is good news that that means our installs went correctly but yes we will in fact be doing something a little more exciting than that we will write a really quick program so we have installed the libraries we have loaded the libraries now we have to create an object or an image recognition network okay and so we're just going to call it net and this is where we're gonna use that training model that we downloaded here in those last steps so how do we create our network well I'm going to name it net you can name yours whatever you want but I'm gonna call my net and then it's going to go out to Jetson dot inference that's that library that we just loaded and then it is going to do a detect net okay now we have to tell it which network model to use now remember we downloaded them all so there's a lot of different image identification networks that our training data that we trained trained models that we downloaded and so the one we're going to use is SSD - mobile net - the - okay and then we need to set a threshold okay is equal to 0.5 and so if you set the threshold low it's gonna recognize a lot of stuff if you set the threshold high it's not gonna recognize much stuff and so let's say that I have a pin that's wrong not a good example let's say I have a cup okay if I set the threshold to 0.1 it's gonna always recognize the cup but it might not recognize the microphone as a cup if I put the threshold at 99 or 0.99 then you know it would not have any false positives but it might miss some cups and so about a point five is not perfect but usually most things that we'll get right you can play around with that threshold after we get the program written all right now one of the things that you're going to see now that we're using these jets and utilities we're not going to do the CV to create a camera using CV - but what we're gonna you know how we created those video capture objects we're going to create our camera object using the jets and utilities so what I'm going to do is I'm going to say my cam is going to be equal to jet son this is before I think we use like CV to video capture but we're using the jets and utilities now so we're gonna say Jetson dot utils I'm gonna create a GST camera and now I need to give it the resolution well I've already looked this up so I know that my resolution is 640 by 480 will work and then I'm going to use my webcam which is going to be slash de vie slash video 1 and if you were using the Raspberry Pi camera then you would need to use that cam set you know that long string that we use to create a video capture object well you would need to use that long cam set string here but we indeed are using the webcam and so I can just use this let me kind of show you that you've got to make sure that this you've got to make sure that this resolution that you set is one that is supported by the camera and so I am going to run over here and get a useful command that's not the right thing I keep useful commands over here and this is the command remember we installed VL 4 and so now I will come back to a terminal and now I will paste and then I'll explain what I'm doing ok and I will in fact get out of your way all right so we give the command V for l2v for l2 - you know what I'm going to do I'm sorry I'm going to change directory back to my home so that you can see more the line easier and now I will explain the command you do v4l to - CTL and then you got to tell it this is like saying tell me what the camera parameters are and then you've got to do a minus D which says which device are you looking for well slash dev slash video zero would be looking at my PI cam because that's zero but this doesn't work with the PI camp so I'm gonna go to camera one which is my USB camera ok so - d what device are you looking at you put a spat space and then /de V / video 1 so for me that's saying that that's going to be my webcam and then space - - list - formats - e xt and so this should show me my webcam ok very good and do you see all these different sizes that it says I can do it shows me the sizes and then it shows me the frame rates that it's capable of okay and then here is a 640 by 360 and you see you got up here at the top 640 by 480 so I'm gonna go ahead and set that to 640 by 480 as the size but you should be able to use any of those sizes that showed up here under your camera now if you have two webcams you could also look at /de v / video - okay so you can get all your different sizes there that you can use now the one other thing that I will show you is I don't think that you will get much for me the Raspberry Pi cameras video 0 and so this probably won't say very much yeah and it just this is all the information that it gives you but those are the resolutions so that seems like that would actually work for the Raspberry Pi camera as well that command so and I do believe that we are ready to jump in and start coding one of the things that you can just see is is that if we are I'll just make note of this right now because I know where we're going that if we're going to be using this video 1 which is going to be my webcam video capture one that I can use 640 is 4 and 480 as one of the allowed sizes so that's one of the things that we'll need to know but let's go ahead and fire up our visual studio code and you can see that I have created a new folder which I call Nvidia and then inside this Nvidia inside this Nvidia folder I'm going to create a new file and that new file is going to be called my identify - 1 dot pi you can call yours whatever you want that's what I'm going to be calling mine the good news is I know these fonts are hard to read over here but I do have a big enough font that you should be able to follow along by follow along with me now this is going to be just a few lines of code but it's gonna be a pretty impressive demo so let's get ready to roll here so the first thing we're going to do is we spent a lot of time installing these new libraries we're going to a couple of them we are going to import Jetson dot inference what does Jetson dot inference do that is the program that'll that is the library that will allow us to look at a frame and find things of interest in the frame sort of like our face recognition library was in our earlier programs then we are going to import some utilities Jetson dot utils that library allows us to interact with our display and it allows us to interact with our camera okay with those things out of the way we are going to need to do a certain a couple of things here give me just a second here I need to get a little bookkeeping done over here that looks good okay now with those two pros two libraries in we can go ahead and create our camera you can remember that when we were doing OpenCV we did like a video capture type command cv2 video capture well now we're gonna be using the Jetson utilities to control the camera and so we are going to go ahead and setup the camera as Jetson dot utils UTI LS Jetson dot utils and then we are going to do GST GST camera so you see GST uppercase C for camera and now we need to give it a width so we're going to give it 640 for the width and 480 for the height and then we got to tell it which device and I'm going to put a dev a slash dev and then slash vid the video 1 and that is my web camera all right now if you are gonna want to just use your Raspberry Pi camera all you would have to do is right here you would just tell it that it's camera zero so you see the the PI cameras are a little bit different beast than the webcams and you use a kind of it like a different way of interacting with them so if you're going to use your Raspberry Pi camera you just put a zero here I don't have a dual camera I don't have a dual camera nano I'm wondering if you had two cameras on your nano if the other one would be one that would be what I would guess but I know zero will work for the for the the one camera setup so we will go back because I will be using the USB cam on this so that should set it up now we've got to set up a displace it will say D is P for display and that is going to be Jetson dot utils and then we will make a GL display just setting up that object so now I have a cam and I've got a disp now I need to tell it a font because I'm going to be interacting with it text you'll wise since I need to say the font is Jetson dot utils dot coud font okay and there's really not any options that I'm aware of you just kind of get the font that you get which is the CUDA font so now I have a camera set up I've got my displace set up and I've got my font set up you can see this is a lot like the kind of mindset that we had in open CV it's just our commands are different now this is a little bit different too now we're gonna go into that go out grab a frame do something show a frame and before an open CV we would just say while true but here we're gonna be a little bit smarter we're gonna say while okay this pull to display dot is open pay attention to the upper case lower case display all lower case dot is with an upper case eye and open with an upper case Oh is open alright so this just means we're gonna fire off a window and as long as that window is open the program will run but when we kill the the window then it will kill the program and exit us out and release the camera and that sort of stuff so it's it sounded kind of like now we're not going to have to have to do that wait key business so now we're gonna go out and we're going to get a frame and then we're also going to get returned to us the width of the frame and the height of the frame and so now it seems like every time we call something or get something returned we're sending it the width and the height and when we grab the frame it gives it back to us and so this is going to be equal to cam dot capture with an uppercase C RGB a alright so right off the bat a couple of other things you notice different than open CV and open CV we were working in the BG our color split space blue green red here we are working in the red green blue alpha and that fourth channel is one that's just sort of like opacity and transparency so if you want to put a box over someone's face that has some degree of opacity you've got a parameter that you can work with that so cam is cam dot capture rgba so now we have a frame now we got to figure out what does it see in the frame so we're going to do that kind of detection that detection step so what I'm going to say is it's going to tell me the class ID that means it's going to look and if it sees a cup it's going to return to me not the word cup but a number that would correspond to the word cup so it would be so like you've got a list and it's the position in the list so class ID and then also it will tell me how confident it is in its identification and we do that with simply net dot classify so we're gonna net dot classify what we're we're gonna look at the frame and again we got to tell it the width and the height now make sure that you're passing it the width and the height that you got back from the camera because you set the camera up as 640 480 but what if you messed up and that wasn't one of the allowed ones and therefore it set it to something close nearby so you want to make sure that you might have asked for this but you want to make sure that you pass it what it told you it was so let's say if I said sick forty by four eighty one I wanted it might return to me an image that is 640 by 480 well we want to use that with it height here to make sure that things match up I hope that makes sense I hope that's not too confusing so now I have the class ID so let's say that I could do cup pen pencil well it might return the number of the class ID 1 which would be cup or two which would be pin or 3 which would be pencil so once we get that class ID it's a number now we got to go look up in a list to see what the word is what the actual object is that goes with that ID and we do that rather simply we can just say the actual item that we found was a net dot get class description upper case G uppercase C upper case D get class description of what we'll just the class ID that I just found all right so now I should actually know what my what my item is that I found and now I want to put that on the frame so I'm gonna do my font which I've already set up my font dot what overlay text and so this is gonna overlay text overlay it overlay it on what overlay it on my frame that I grabbed anytime I use frame I got to tell it again with height okay and now what do I want to overlay I want to overlay a word well what word do I want to overlay item whatever that item was that I found so I'm gonna be labeling that frame with the word item which might be cup or pencil or pen or whatever and now I don't want it up in the very top corner so I'm gonna come down about five and scoot over about five so I'm gonna kind of give it a x and a y offset of about five pixels so I'm not jammed all the way in the cup corner now what color do I want the font well I'm gonna say I want it fought magenta so the text is going to be magenta and then what do I want the background the background of the the letters I'll make it font dot blue and the obvious colors it will know like this okay now I've got the image I know what is in the image and I have overlaid the frame with the word what it is now I just got a show so remember I'd do an IM show in CV - well instead of I am show I will do a and this was dis my display object up here it will be disp render once and then what am I gonna render I'm gonna render the frame and remember that frame has the text on it and again you always tell it what and height okay so that's a pretty simple little program and even though all these commands are new I think you should kind of recognize what we're doing here so let's go ahead and let's run this thing and see all of our problems now invalid sent it Jetson utils what did I do wrong there that looks so right oh in part i'm not going to impart i'm going to import all right now i'm going to warn you the first time you run this it's going to take five minutes to run because the the tensor and all those types of things are going to go in and optimize and it's going to teach itself how to learn fast i mean it's going to teach itself how to recognize fast so the first time you run it it's going to sit and optimize itself for about five minutes but you're going to see a lot of stuff happening it's not going to hang but it's going to take you a lot longer than me because i've already run this on this machine so let's go ahead and say run Python file in terminal oh it still doesn't like it what does it not like failed to create device we will come up here and I have 640 by 480 and I've got / div / video 1 and that is line 4 utils GST camera can you guys see what I did wrong there 640 by 480 ah I forgot I forgot the /dev okay let me show you you've got to go back to roots / DV / vid D video 1 right you've got to go all the way back to the roof you got to have that leading slash and this time I think it should run it's looking good looking good what did it not like that time 1 8 display ah I didn't call it display I called disp right so when I named my object up here I called it disp so I gotta call it disp their run Python file in terminal okay it did not like that net is not defined ah I didn't set up my net that was not good okay so what object detection Network are we going to use I've got to set that up net is going to be equal to Jetson dot inference dot image net so I got to tell it what training set what network what you know what model I'm going to use to recognize images and remember we downloaded like a dozen of them well the one we are going to use is Google net now you can go in and use the other ones and to kind of play around with the the other ones but they need to be the image net models not the detect net models this program will work with the different image net models so you can try the different ones and you can go back and look and see the ones that we have okay so I am going to come in now and I am going to run this thing and this is where it's going to take you five minutes right it's going to take five minutes for you to optimize but I've run this before so it'll just take maybe 30 seconds but this TRT is gonna take a while for you to optimize okay well look at that right off the bat boom look at that this is my green screen back behind me but what does it think it thinks the green screen is a shower curtain well I would say that's a pretty darn good guess and now let's just see if I put this up here look at that a pill bottle got it right off the bat giddyup that was pretty impressive wasn't it what if I put this it's gonna recognize that as it says a goblet I guess that would be a goblet that's is a coffee cup you might ask me why is my coffee cup so dirty well it's because I like to season my coffee cups and so I do not wash them it actually enhances the flavor in my mind okay let's see what we can do here drumstick paintbrush okay ballpoint pen there it got it ballpoint pen that it is pretty darn good let's see what other things I've got here I've got a camera let me turn the camera on okay and then if I come here let's see if it can recognize it as a camera it had it there for a second I think this will work better if I point down and give it a better view okay so let's give it a reflex camera there it got it pretty good just needed the right view I come over here mouse computer mouse I come over here and I've got a it looks at it as a whole notebook but it's got the computer keyboard that is pretty good look at that it X actually recognizes the spacebar keyboard okay that's pretty good computer keyboard look up here it sees it as a monitor wow this is pretty impressive and let's see if I can you see it sees my lighting system is an umbrella over here it sees that I've got a desk over here I'll look at that it saw the soap dispenser on the wall and then it thinks of it is a rocking chair let me see if I can give it a little better I'm not sure why it sees that as a rocking you're a rocker that certainly doesn't look like a rocking chair but let's see if I can give it a little different view there okay now folding chair is a little bit better of a guess and so this thing you can see just looking around the room it's able to recognize quite a bit of different stuff desktop computer again desk let's just see if I just pointed out there it sees a lot of desks out there that's pretty good now it sees a bunch of books over there on the thing and so it was kind of guessing library can't quite see my airplane all the way back in the back so anyway you can see that this thing is finding a lot of different stuff and we're going to spend some time you're really going to learn a lot about what is going on with this software but I mean with this with this library but I just I just think it's pretty neat that in just a few minutes we were able to write our first image recognition program and so this is kind of what we're gonna be doing moving forward in the future is learning more and more about how to identify objects and so that was just sort of a quick demo and we will be going into this type of thing in more detail in future lessons okay this was pretty much fun I enjoy getting the program written it was really a lot of work on the installations and I felt like we should at least have a little bit of fun at the end of the at the end of the lesson so next week we'll come back and we'll really start developing this a lot more with this with this library plus I promise you next week I'll have a better way for you to see which libraries are at your disposal okay guys if you like this video give me a thumbs up remember to subscribe to the channel when you subscribe make sure you ring that Bell so you'll get notifications when my future lessons come out and I will talk to you guys later

Channel: Paul McWhorter

Views: 10,935

Keywords: NVIDIA, Deep Learning, DNN

Id: 5rbOsKCZ-VU

Length: 48min 51sec (2931 seconds)

Published: Sat Jul 04 2020