Multi-Object Tracking with Ultralytics YOLO | YOLOv8 tracking
HTML-код
- Опубликовано: 5 сен 2024
- Object tracking is the intricate process of recognizing and following the movement of specific objects, be it one or multiple, within a sequence of frames across a period of time. This involves continuously pinpointing the object's location in each frame and maintaining a record of its trajectory as it progresses through subsequent frames. The versatility of object tracking finds applications across diverse domains, including surveillance, robotics, autonomous driving, sports analytics, and more.
Github: github.com/Aar...
For queries: You can comment in comment section or you can email me at aarohisingla1987@gmail.com
A renowned solution for object detection and tracking is YOLOv8, which provides a user-friendly implementation of tracking algorithms through two distinct architectures, currently:
1. BoT-SORT
2. ByteTrack
Both BoT-SORT and ByteTrack furnish robust and efficient tracking algorithms that can be seamlessly integrated into your computer vision projects. While the default tracker is BoT-SORT, you have the flexibility to opt for ByteTrack if it better suits your requirements.
#yolov8 #yolo #ultralytics #computervision #objecttracking #objectdetection
I recently asked a question about ”YOLO-NAS Pose Implementation”. I understand that combining pose estimation with tracking enables us to obtain information about a specific person. This insight is incredibly helpful. Thank you so much for providing such valuable information.
You're very welcome! I'm glad to hear that the information was valuable to you.
Hello Aarohi
Your channel is very knowledgeable & helpful for all Artificial Intelligence/ Data Scientist Professionals. Stay blessed & keep sharing such a good content.
Thank you, I will
Great explanation thanks
Glad it was helpful!
@@CodeWithAarohi from your point of view which is better supervision and byteTrack or ultralytics track command with byteTrack
Spectacular content . Please keep uploading awesome like this.!!!
Thank you, I will
@@CodeWithAarohi will you help me ?? cant able to track i can share my code
Amazing work done. Keep it up
Thank you, I will
Very nice
My pleasure 😊
Excellent tutorial to get an idea on yolo. Simple and sweet explanation. I tried the example but as soon as the model.track is executed I'm getting 'Kernel died and will try to restart automatically'. Runs fine if I remove save=True. Any idea what causes this ?
try to reduce batch size and image size and then check.
how to track accurately overlapping objects ?. (because this method doesnt work for overlapping objects )
your videos and work are amazing! i have one question can i use pretrained model on coco clases for detect only one class for example just car, i dont know how to do that and is it posible? thanks you
Yes you can.Just pass the classes parameter like I have shown here : results = model.track(source="1.mp4", classes=3, conf=0.7, iou=0.5, save=True)
3 is the class id of car class.
Ma'am, Please make tutorials on DeepStream SDK
Sure
hello I have a question I have everything like you and working good but they Keypoint I can't rename them can you tell me how can I rename the Keypoint thank you
Can we get the tracking position for each and every object? And define a unique target position for every object we are detecting?
Thank you
You're welcome
Hello, This video is very helpful, by the way, I want to train my custom model based on single class. Can you make a video on that where we can train our custom model with the custom data ? where you can explain all the parts that needs to be done in order to do that. from dataset generation to end-to-end.
Sure
Aarohi , Yolo v8 comes with multiple tracking options and you also mentioned few additional ones, whats the criteria that one should use to decide on which tracking algo to use . Do you have a video on that ? Maybe i missed that , can u please share the link
No, I don't have a video on that but will try to make one.
Hi, Thanks for the information.
Can you generate a confusion matrix for test data (detection task)?
Can u explain the configuration of bot-sort tracker
Sure
Why is the ID number always updated? I want to keep the ID number constant in every frame, how can I do it?
Check this recent video - ruclips.net/video/Z3uquMElyzI/видео.html
Can you make yolov8 weapon detection ? I was wondering to use yolov8 and CNN to making weapon detection
For that you need to train your yolov8 model on weapons and then use the best.pt model as I have used here.
@@CodeWithAarohi do you use CNN method?
mam i want to connect with you on linkdin could you please give me your profile link?
www.linkedin.com/in/aarohi-singla-761b2448/
where is the mp4?
It was not there. You can download any video from internet and test.