Yolov8 object tracking 100% native | Object detection with Python | Computer vision tutorial
Вставка
- Опубліковано 20 сер 2023
- Code: github.com/computervisioneng/...
🌍 Community 👥
Join our Discord server: / discord
Support me on Patreon: / computervisionengineer
#python #computervision #yolov8 #objectdetection #objecttracking
Dive into a world of exclusive computer vision content on my Patreon: www.patreon.com/ComputerVisionEngineer
Is there a better library with lots of objects like in the thousands or tens of thousands, and it will also recursively tell you the features within objects. I want to use this to build an immersion environment for learning Chinese, I want to use it on live streams of people playing video games, and movies.
Ive never heard someone narrarate like this but it actually makes it way easier for to retain information. The exaggerations on the important things help it stick. Some tutorials people are monotone and talk so much about unrelated things that its hard to know whats critical and whats not cause its just peppered inbetween a bunch of droning. I can listen to this guy at work and know whats key by his tone of voice and dont have to dedicate myself to following a video down to the T cause i dont know what tiny detail will cause the whole thing to not work.
Agreed. Exaggerations are a very useful teaching resource. Glad you enjoyed it! 😃🙌
Borat style
Nobody f$&@S with the Jesus
Your tutorials are go great. I was able to complete my first computer vision project because of you. Thank you!
Just want to say that your tutorial and the files provided are very helpful and convenient. Big thanks to you!
Thank you! Glad you enjoyed it! 😃🙌
can I ask u what was your dataset pleaze
This is exactly what I needed!!! First test that perfectly worked with my own custom model and dataset. Thank you for the brilliant work!
You are welcome! Glad the video was helpful! 😃🙌
This long project in such less duration and that too so easy to follow and engaging at the same time. Hats off! 💯
Thank you once again! 😃
Very useful, I was able to repeat the steps and it worked just fine!
Please help!
I have multiple classes of object detection model, but I don't know how to use the detection results to control an arduino board. I've stuck with it for a week.
Ahhh after mindless searching finally this is the code been looking for ❤
Simple and Excellent tutorial. Thanks.
Thank you for the amazing video!
I see you have now started showing how you created the project in PyCharm as I asked for; You are really a Blessing and Heaven-sent!!! GOD Bless you richly!
Absolutely! 😊 Thank you for your good wishes! 😃🙌
Hi, I collected data with video to use in my project. Can I use this video to train my model or do I need to train the model using only photos? If I can use it, how should I label the data in this video? I would appreciate it very much if you could help me on this issue.
how do I export a file that contain information of detected polygon coordinates and the frame index at which they are detected?
You are insane!!! 💯 Amazing explanation 👉👈
😃 Thank you!! Glad you enjoyed it! 💪💪
Hi. After giving persist=True, till how many frames or how long does it persist
How can I find a ready-made template so I can count the number of people for my project?
Is it possible to detect a single particular object on the screen by selecting that particular bounding box (or creating a bounding box around the object using mouse)? And then it can track only that particular selected object. I am trying to create a model for single object tracking using drone. Please let me know if anyone has any idea?
Great work as always, mate! Saving this to watch later!
Quick question-can you track the coordinates of the bounding boxes?
yes. You can then use the center of the rectangle to draw a line between frames to show trajectory.
Thank you! 😃Tracking the coordinates do you mean drawing a trajectory as @AlainPilon said?
@@ComputerVisionEngineer That was exactly my question :) Thank you both!
Im a beginner. your explanation is superb😍
Thank you!! Glad you enjoyed it! 😃
Hi! How we can use it to track the keypoints of objects? Is it possible?
Some time track I'd is changed for same object. Any solution?
Well done and easy to learn
does it work with occlusion for long range tracking ?
Really like your style and explanation. Kindly upload full detailed videos. Thank you
Thank you! Glad you enjoyed it! 😃 I will make full detailed videos! 💪
your video is good, i ran it working fine for first frame, but for next frame it exits and kernel dies out. Can someone pls help me
Hi!! i love your videos they are very very helpful!!!! - quick question: this method outperform traditional tracking (SORT -DEEPSORT)? Thank you! greetings from Chile!
Thank you for your support! No, I don't think it outperforms other object tracking methods.
thanks alot, my job is going somewhere now.
Thank you for your video, its better if you show which properties in "result" variable covers tracking parameters like ID, Class, Box.
Hi! Your videos are really helpful, so thank you so much
i have a quesion how can i detect, track and count at the same time
please i need the repo of deep sort that is compatible with the last version of python
hey felipe, i wanted to ask whether python 3.12 is required for yolov8 ? because i am getting error by using python 3.8
I have used Python 3.8 with no problem. What error are you getting?
This is like getting a programming lesson from Frenchy from The Boys. I love it.
Also.. super helpful... and fml, had python 3.12 and couldn't get yolo to install because pytorch won't run on anything higher than 3.11 at time of me writing this.
Frenchy from The Boys 😂😂
Cool, thank you for your contribution regarding the Python version! 🙌
Thanks for the Brilliant video! Do you know of any ways that the plotting time could be sped up so as it calculates the video at near to real time?
🤔 Using a gpu would speed up the inference. About speeding up the drawing, perhaps using c++ instead of Python. You could also do the object detection + tracking on 1 every 5 frames, and interpolate the bounding boxes for the other frames. 🙌
@@ComputerVisionEngineer Thanks so much. Your really the best AI coding channel on UA-cam.
Thank you so much for your guides!
Would be nice if you show how after this steps get exact objects, fe only persons or only dogs, also take coords of objects :)
You are welcome! Sure! I will keep it in mind in future videos! 😃🙌
hey, great work as always, Quick question: I am using this method to track skin lesions on my face, I wonder if I am able to get the exact coordinates of these tracked points? :D
Hey, do you mean the exact coordinates of the bounding box you are tracking through the video?
@@ComputerVisionEngineer yess, exactly ;D
Beautiful, excellent, intelligent, extraordinary ❤❤❤
Thank you for your support! 😊🙌
great video
If I want to track only the person class, then what is the default variable name for predictions? like class_id or pred?
results = model.track(frame, persist=True, classes=0)
How can I create the model to just detect the humans and not other things?
Take a look at the ultralytics documentation, you can specify what are the classes you want to detect.
Thank you for this amazing guide. I'd to ask, please how do you set ultralytics on Vscode? I usually run into an error when I use pip install ultralytics
You are welcome! Not sure if I would be able to guide you on how to install it in vscode, I use pycharm. 🙌
@@ComputerVisionEngineer I used pycharm and it ran, but I'm getting no output
Thank you for share knowledge. It's very useful. It simple but powerful code.
Agreed. Simple but powerful code! 💪💪
thank you for the tutorial🙏 i have a question, is it faster to use yolov8 native object tracking or deepsort?
🤔 I think yolov8 native tracking is faster
Thanks for tutorial. it is great, but I keep getting "source is missing" error
Hola buenas, ya que supongo que hablas en español te hablo en español jaja, una consulta, si en ves de un video tengo todos los frames de ese video, que debería cambiar en relación al código del video? Saludos desde Chile.
Hola, si tienes todos los frames guardados en un directorio por ejemplo?
for image_path in os.listdir(frames_dir):
frame = cv2.imread(os.path.join(frames_dir, image_path))
....
hi sir , in parking space detection video , how can i numbering the parking spots like that 1 2 3 next to each parking spot.
thank you for works
Hey, you can write the numbers with cv2.putText, and should be able to get the id for each parking spot from the 'connected_components' object. Take a look at the line 31 from util.py file. 🙌
Hi, how is this version different from the video which you used deep sort?
Hi, in this video we are using yolov8 only to do both detection and tracking. In my previous video on object detection + tracking using deep sort we used an external library (deep sort). 🙌
AttributeError: 'YOLO' object has no attribute 'track'. Did you mean: 'task'?
How can I contact you for private consultation?
did you detect and track in only One sentence, That's amazing, i don't know how did you do that.
Yes, agreed, yolov8 is amazing! 😃🙌
Cool👍👍👍👍👍👍
this is cool
Nice video sir ❤
Glad you enjoyed it! 😄🙌
I might have missed: where can I download the video? I would like to follow your exact foot steps
I have added a link to the video in the readme file of this tutorial's github repository. 🙌
Is there any tutorial to track and show bounding box of a object with specific ID?
Person Identification and re-identification is still a research area. there are many research papers regarding, but no much resourceful implementation
If you have any resource, lemme know too
You are a World referent in IA, thanks
Thank you so much for your kind words! 😊🙌
i wanna dedect car make model car how i can do
Just wanted to know If I want to detect only a person from this video, how can we this?
Do you mean if you only want to detect the class 'person'? You can specify what are the categories you want to detect when making inferences. 🙌
in ultralytics package i can already use my cuda gpu? not cpu?
yes, you can use your gpu
you may have to uninstall the preinstalled torch version and update it using a torch update compatible with your system + CUDA version
Thank you very much , i try to test object tracking but error , i found ans from your vedio
Why don't you use "SORT" algortithm for tracking as in other videos?
No reason. It is always healthy to get familiar with different technologies in order to solve a problem. Yolov8 built in object tracking is very easy to use and a very interesting tool in order to do object tracking. 🙌
100000000000000000 likes excellent explanation
Thank you so much for your support!! 😊 So glad you enjoyed it! 💪
i want to make a software of face attendance for my company with professionell GUI can u do project like this
I will try to. 🙌
how to save the data that you display to one whole video ?
Do you mean how to save a video to disk with the bounding boxes on top instead of showing the frames one at the time? Tale a look at this script github.com/computervisioneng/train-yolov8-custom-dataset-step-by-step-guide/blob/master/local_env/predict_video.py 🙌
Do it with counting how many objects cross a line in the frame
I will try to. 🙌
Great tutorial. For some reason its not compiling for me. I think im making a mistake with the path. New to python and not a great programmer lol
Thank you! Maybe someone in our community can help you. Try asking in our discord.
@@ComputerVisionEngineer i figured it out on my own now lol
i like ur accent hhhh understandable more then native one
Thank you! Every new video I try to sound a little more understandable, I also upload subtitles just in case. 😃🙌
I want to build a system of robots by connecting wheels to my phone/tablets, and they will then take pictures of everything in the house and tell me where things are... I wish the robot can charge itself haha.
Sounds like a cool project!
chilean?
Uruguayan 🇺🇾
short video, less explanation, lot of abstractions (what is this what is that?). Not recommended for beginners. Go find another youtube tutorial.
Muh se Supari nikal kr baat kr ray baba ....
Hey Felipe when I tried to run your code, I got an error at results = model.track(persists=True). The issue was lap >= 0.4 not found and the update was skipped. Do you have any idea how to fix this error?
Hi
I have the same problem😢
How did you solv yours??
Good Morning, I try to use the same code but instead of 10 frames I use the whole video,
The code runs ok for multiple of video frames and after some time I get this particular error:
Traceback (most recent call last):
File "c:\Users\user\Desktop\detect1\main.py", line 38, in
track_ids = cont_tracker.update(np.asarray(detections_)) # contains boudning boxes and tracking related information
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\user\Desktop\detect1\sort\sort.py", line 232, in update
matched, unmatched_dets, unmatched_trks = associate_detections_to_trackers(dets,trks, self.iou_threshold)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\user\Desktop\detect1\sort\sort.py", line 163, in associate_detections_to_trackers
iou_matrix = iou_batch(detections, trackers)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "c:\Users\user\Desktop\detect1\sort\sort.py", line 55, in iou_batch
yy1 = np.maximum(bb_test[..., 1], bb_gt[..., 1])
~~~~~~~^^^^^^^^
IndexError: index 1 is out of bounds for axis 1 with size 1
I would be grateful if you could help me resolve this