YOLOv8 for Beginners: A Complete Tutorial on Windows 11

I'm going to show you how to run yellow V8 in under 20 lines of code now not only will you be able to run yellow V8 but you'll also be able to run yellow V7 your V6 your r x and Beyond you just change one flag in your code and you'll be able to adapt to any two versions of YOLO which is super cool if you go into github.com augmented startups as1 right and this is essentially a modular library for all YOLO object detection and tracking algorithms and this means that you can use it for detection and tracking so looking over here we are able to use deep sort by track and nowhere so say for example you want to use yellow V5 but you only want to use it with no fear but later down the line you decide oh no I don't want to use your level five maybe let me use test.iola V6 then maybe you can say okay let's have yellow V6 Nano with no fear but then you realize later on that you don't like nofe and maybe you want to go back to deep sword and you can do this All by changing a single line of code in your project and if you scroll all the way down you can use YOLO V8 right off the bat so let me give you a a typical scenario of why you'd need this because say for example you are using yellow V8 right and you built your whole project on it and you know things are working really great but all of a sudden someone comes up with the next version of YOLO which is Yolo v9 which is so much better than yellow V8 then you'll be like oh damn I need to change my whole project code just to accommodate the next version of YOLO not to this library because if next version of YOLO comes out you just need to change a single line of code as I've mentioned earlier so you change this from yellow V8 to yellow v9 or yellow V10 whatever feature version of YOLO comes out and instantly you would be able to continue with your project and you don't have to change anything else isn't that so cool isn't that so cool right so let's get started with it let me show you how it's done so we're going to go into as1 make sure you have a folder so this is on your PC you can create a nice folder I call it yolo now I'm going to open up my command prompt CMD right and now what we're going going to do we're going to scroll down here to the instructions and we're going to git clone this repository so now we are cloning the as1 repository right as you can see the installation is super simple right now we've got that we're going to create an environment a new virtual environment so let's place this over here and while that is there let's move this side by side with this right now we've created the environment now we can activate our environment with this cool now we know that our environment has been activated because we have this dot end over there next we're going to install siphon cool so everything worked out perfectly and if it prompts you to upgrade and now you can copy this thing right over here Ctrl C Ctrl V and you can upgrade pop now it's not completely necessary I don't know I don't like seeing these yellow warnings over here so I decided okay let me do it and then while that is getting ready let's go over here and copy this thing over here so this is siphon B box so that's the bounding box for that we can use with Satan okay so we're going to copy that paste it over here and enter cool now that has been completed next we're gonna pop install as1 so we're gonna copy that place it over here and enter now this will install all of the libraries that you need in order to run as1 so while as1 is installing the next step that we're going to do is to install Dodge Vision now if you don't have a GPU you're going to use this command over it so that you can run it on CPU but I highly recommend that you install it on a GPU because you'll get real-time performance on as1 not only that but for any computer vision algorithm it will always run faster and real time on GPU rather than on CPU cool now that we have as1 installed I'm going to copy this one over here for GPU copy that paste it over here and let's run it so torch Vision has been installed now is the moment of truth let's see if everything works now don't forget that we need to CD into S1 so that's over there okay if you're going to our YOLO folder we'll see that we have as1 over here now I made a mistake right over here this end supposed to be inside here but let's copy all of this cut it out and put it right over here okay normally you'd first have to change directory and then run all of the installation instructions so keep that in mind cool but it's not a stream smash let's see if we can run our video so let's python main.bi data sample videos test dot MP4 so let's run it and see if everything works so I made another mistake again so I need to go out of it so change directly back into the previous directory and let's run it again okay cross fingers cross fingers let's hope it works so right now it's downloading the YOLO V7 model we'll change it to Yellow V8 in just a minute but let's see what happens look at that it works and not only do we have yellow V7 working it's also working alongside a Tracker now let's find out which tracker we are using exactly so I'm just going to Ctrl C out of this we're going to go into our yellow folder that we had earlier right and we can delete this one here this is one the mistake that we made now what has happened is that when we ran the the code it downloaded all of the stuff that we needed for your V7 to run now I'm going to go into main.pi right let's open it up and then while we are here and waiting for it we are also going to go into benchmarks and this is where we're going to get the flag that we need for yellow V8 so now we have main the pi and as you can see we have our detector over here we have our tracker so I'm going to change yellow V7 by torch let's do something like Iola V8 medium and we're going to choose the Onyx model right let's just select that and paste it over there let's save it and let's run it again so I'm going to say python main.pi everything is just the same and we're going to test it on the same video now if you want to change the video you can go over here too data sample videos and we've got a wide variety of videos that we can use right off the bat so as you can see right now it's downloading the yellow V8 media model and it's running look at that it's working really well there is a false positive which is this one over here we didn't see that in yellow V7 but for some reason it's picking up in the lv8 now we can adjust the confidence threshold to filter out some of these false detections but also you can play around with the larger models which will give you better accuracy now before we go I want to show you how we can swap out bike track object tracker with something else let's change it to no fair so we're going to try out model flag no fear we're going to put that over here paste that there and let's run it again python may not buy data sample videos this dot MP4 we've got our tracker running everything looks hunky-dory and yeah look at that it's running quite quite nicely now just note that it is running a bit slower on my system because I am simultaneously recording this video and also we are running object tracker in conjunction so now it will run between 12 to 15 frames per second cool so once you have this up and running and you've played around with the different YOLO models what you can do now is to head over to store dot augmented startups .com and here you'll find a wide variety of projects that you can try out for yourself so right now we are adapting all of the yellow projects to be as1 compatible so you can run real-time read detection in under 30 lines of code you can run Mass detection in maybe 20 lines of code and so on and so forth traffic light detection maybe we could do that in 50 lines of code Apple Center Stage that will also be another one that will be converting so you'll be able to run traffic detection with color recognition using yellow V7 V8 R and so on and so forth isn't it really cool one library for all your project needs so we will be updating this very soon so make sure that you stay tuned to augmented startups.com and we'll be bringing you all of these projects very very soon now I know I said that we will do this in a 20 lines of code and right now it's showing under 50 lines of code I mean look at this this is very minimal lines of code like if you take this out like we don't really need our arguments we can condense this into one line we can condense this one also to one line and if you take all of this out we will definitely get to under 20 lines of code isn't that amazing and one more thing if you want to run this in Google collab then there will be a link to the video right up here

Channel: Augmented AI

Views: 13,136

Keywords: YOLOv8, Object Detection, Image Segmentation, Ultralytics, YOLO family, SOTA model, Fast, Accurate, Easy to use, Training models, Instance Segmentation, Image Classification, Cutting-edge, Performance boost, Flexibility, Previous YOLO versions, New features, Improvements, Wide range of applications, Fast object detection, High accuracy, Unified framework, State-of-the-art, Object recognition, Image analysis, Machine learning, Computer vision, YOLOv8 tutorial, LearnOpenCV

Id: C4i8Bia83TM

Length: 8min 59sec (539 seconds)

Published: Mon Jan 30 2023