search

roboflow / sports

computer vision and sports
MIT License
2.4k stars 256 forks source link

Include ball in radar view? #10

Open gdubya opened 3 months ago

gdubya commented 3 months ago

Great project! Thanks for sharing. I hope to test it out more extensively in the near future.

Is it possible to include the ball position in the radar view? That would give a bit more context to the player movements.

SkalskiP commented 3 months ago

Hi @gdubya 👋🏻 yup, I will probably include it in the radar view.

druefena commented 2 months ago

@SkalskiP Note that including the ball into the radar view is tricky as it is not always on the ground and hence its location will be wrong for flying balls. Any ideas how to tackle this issue?

SkalskiP commented 2 months ago

Hi @druefena 👋🏻, for now, I was just thinking about doing it in a naive way. But you're absolutely right. Do you have any ideas on how to tackle this problem?

ms82119 commented 2 months ago

@SkalskiP Watching the circles that represent the ball position over time, you can often see the parabolic trajectory. Perhaps there is a way to look at this history to determine if the ball was in the air. It would likely need to be done as a post processing step as you would need to get the ball path coord history first and then to try and estimate the height... seems a bit complex though as the camera is often moving as the ball is!

Btw, thanks for posting this repo, you've combined a lot of separate concepts into a really useful and practical repo. I'm learning a lot from it :)

fenaux commented 1 month ago

@SkalskiP @gdubya @ms82119 @druefena Please consider the solution proposed here https://github.com/fenaux/soccer-applications

clairesecehDS commented 3 days ago

@fenaux, your work on that is precious; thanks a lot! But I have an issue: the ball is poorly detected in front of the stands. So, the parabolic trajectory is uncompleted. Do you have the same problem?

fenaux commented 3 days ago

@clairesecehDS on which video ? Due to the fact missing detections are replaced by a linear interpolation, if that happens during a parabola, this will lead to false linear segment or false acceleration peak. If you wish to submit a video please provide a link in my repo. Thanks for your interest.