predict next bbox

[concrete13377]

Hello! Thank you for your notebook https://github.com/roboflow/notebooks/blob/main/notebooks/how-to-track-football-players.ipynb
Could you please help me understand if it is possible to predict bbox with tracker if detection is missing for 1 frame? For example model detected successfully object on frames 1 2 3 4 5, but on frame 6 model failed to detect it and gave 0 detections. Can I get the estimation of bbox of this object from Bytetrack based on its internal state? What is the best way to do that without just linearly interpolating bbox coordinates?

[github-actions[bot]]

👋 Hello @concrete13377, thank you for leaving an issue on Roboflow Notebooks.

🐞 Bug reports

If you are filing a bug report, please be as detailed as possible. This will help us more easily diagnose and resolve the problem you are facing. To learn more about contributing, check out our Contributing Guidelines.

If you require support with custom code that is not part of Roboflow Notebooks, please reach out on the Roboflow Forum or on the GitHub Discussions page associated with this repository.

💬 Get in touch

Do you have more questions about Roboflow that we haven't responded to yet? Feel free to ask them on the Roboflow Discuss forum. Our developer advocates and community team actively respond to questions there.

To ask questions about Notebooks, head over to the GitHub Discussions section of this repository.

[concrete13377]

Also, I don't understand this part in the notebook. Shouldn't you use the tracks from the tracker as the final detections?

[SkalskiP]

Hi, @concrete13377 👋🏻 Let's convert that issue into a discussion first.

[SkalskiP]

Could you please help me understand if it is possible to predict bbox with a tracker if detection is missing for 1 frame?

Yes, the byte tracker can miss one frame and still maintain the tracker id.

Shouldn't you use the tracks from the tracker as the final detections?

The result form tracker is only box coordinates. If I already have some other metadata (associated with the detection), it will get lost. I need to do the matching to match the tracker id and the rest of my metadata.

[concrete13377]

Can I use ByteTracker not only to maintain id, but to get bbox for frame with no detection? is it possible to prompt tracker to get next "predicted" bbox from it? In the code it seems like tracker can only predict the current frame bbox based on current frame detection

[concrete13377]

by "detection" I mean "bbox from detection model"

[concrete13377]

The result form tracker is only box coordinates. If I already have some other metadata (associated with the detection), it will get lost. I need to do the matching to match the tracker id and the rest of my metadata.

Sure, but coordinates of tracker bboxes and detection bboxes are different, so It seems logical to use coordinates obtained by tracker, and metadata stored in model's detections, am I wrong?

[SkalskiP]

I think it makes sense to use boxes coming detector. Why do you believe tracker boxes are more important? 👀

[concrete13377]

Can I use ByteTracker not only to maintain id, but to get bbox for frame with no detection? is it possible to prompt tracker to get next "predicted" bbox from it? In the code it seems like tracker can only predict the current frame bbox based on current frame detection

ok, thanks! So is it possible to fix missing model detection?