Episode 6 | Pose Estimation with Ultralytics YOLOv8

hey guys welcome to new video in this video here we're gonna do Post estimation with yolo V8 from automatic so the yellow V8 model is also capable of doing post estimation now we can also detect each individual key point on the body we're going to see how to set it up in our own custom python script so this is pretty cool we can use the ulv8 model for so many different kind of things so let's just jump straight into the GitHub repository here where you see all the different tasks we can do with yolo V8 and now we're going to do Post estimation this can be used for making AI trainers motion capture which is actually like a pretty large fill itself so now we can also go down and see the models so we have these detection models we have the exact same ones for post estimations so then we can just go in and directly specify which model we want to use when we're creating instance of the YOLO model so we're not jump straight into Visual Studio code let's see how we can do Post estimation with yolo V8 first of all we need to import our YOLO class from ultralytics we can create an instance of our model now we just need to specify posts if you want to do segmentation we just need to specify SEC if you just want to do update we don't need to specify anything but now we can specify post and then we can do Post automation so this is actually how easy it is to switch between the different kind of like modes or tasks with the YOLO V8 framework so yeah now we have created a YOLO class instance we now have our model we have our post estimation and we are basically going to use the medium model now we can go down and specify the arguments to our model function so we specify the source I have a video here that we're going to throw it through and we can also do it on a live webcam we can set a confidence score if we want to show the results while doing interference we can also save the results so we have this gymnast MP4 so this is the video that we're basically just going to throw through the model so let's just go and run it directly run python file and it should open up in just a second I'm running this on a 49d RTX from Nvidia so it should run fairly fast here we see the inference we can see the image resolution how many persons we're detecting and also the interference speed so this is around 20 milliseconds which is corresponding to around 50 frames per second so here we see that now we're done processing the whole video those number one see the results however runs we have toes and we have predict four and then we can see the video so it's going to open the video in the file explorer so this is actually some pretty good post destination just directly from the exact same model as you can do object segmentation and also object detection from so usually when we do Post estimation and motion capture and so on it is actually like fully dedicated models now we can use your way to perform all these different analyte tasks so let's not go back and run it on a webcam now so I'm just going to specify the source to the webcam and run the program so let's run the program now we should be able to open up our webcam I'm just going to grab the webcam over here there we go now we can see we have the webcam I'm just going to take my arm up here we're going to take the key point but again we should actually see the full body or like at least a large part of the body to be able to do Post estimation here we see that if I just take it up like this we can see it also detects my eyes ears and also the nose so we also get some features and key points in the face so basically get the most important features and key points on our body so just try to leave the camera up here and let's move back there we go so these act like some pretty awesome results as you can see on a live webcam you can also run it on your local machine test it out yourself create some cool projects and applications around it definitely throw it down in the comments here if you have some great ideas what this post information from YOLO V8 can be used for thank you for asking about in this video here I hope you have learned a ton so thanks again for watching hope to see you in the next video as well stay tuned bye for now

Channel: Ultralytics

Views: 15,361

Keywords: Ultralytics, AI, Artificial Intelligence, Object Detection, YOLOv5, YOLOv8

Id: Y28xXQmju64

Length: 3min 50sec (230 seconds)

Published: Wed Aug 16 2023