Object Tracking between API calls

Support YOLO
,
1

Hi Community,

I’m working on an object-tracking system deployed on a server. I developed an API for object detection and tracking using YOLOv8s. An automated script captures frames, sends them to the server, and calls the object detection and tracking API every 60 seconds. However, between API calls, tracking IDs are lost, leading to duplicate object counts.

I’m looking for potential solutions to handle this issue.

Thanks!

2

Hello!

It sounds like you’re doing some exciting work with object tracking! The challenge you’re facing with losing tracking IDs between API calls is common when dealing with intermittent data streams. Here are a few suggestions to help maintain consistent tracking:

  1. Persistent State Management: Consider implementing a state management system on your server to store and update tracking information between API calls. This could involve maintaining a database or in-memory data structure that keeps track of object IDs and their last known positions.

  2. Track Continuation: Use a unique identifier for each object that persists across API calls. You can achieve this by associating each detected object with a unique ID and updating this ID in your state management system.

  3. Increase API Call Frequency: If feasible, reducing the interval between API calls can help maintain continuity in tracking, as shorter gaps mean less chance for objects to move significantly.

  4. Kalman Filters or Similar Algorithms: Implementing algorithms like Kalman filters can help predict the next position of an object based on its previous trajectory, which can be useful when there’s a gap in data.

For more detailed guidance, you might find the Ultralytics YOLO tracking documentation helpful. It covers various tracking configurations and techniques that could be beneficial for your setup.

Feel free to reach out if you have more questions. Happy coding! :blush:

3

Please share the code you’re using to run inference with YOLOv8