tennis-count
Counting scores of tennis balls
1. Counting tennis balls from input: videos of tennis matches.
2. Output: Each player score
3. Steps and literature
https://gml16.github.io/projects/Report_LV8_Project.pdf
4. Action Recognition: https://medium.com/bakken-b%C3%A6ck/improving-your-tennis-game-with-computer-vision-863969743024
5. Get tennis court from videos
6. Track ball from tennis videos
7. Track hit action --> Track ball trajectory --> Track bounce (position of bounce)
--> get the position of the ball inside the court.
8. Track Tennis Ball: https://nol.cs.nctu.edu.tw:234/open-source/TrackNet/
9. Official TrackNet V2: https://nol.cs.nctu.edu.tw:234/open-source/TrackNetv2
10. TrackNet: https://nol.cs.nctu.edu.tw/ndo3je6av9/
11. https://medium.com/bakken-b%C3%A6ck/improving-your-tennis-game-with-computer-vision-863969743024 for Tracking OpenPose
12. Positions of players around the net: https://github.com/sethah/deeptennis
13. The stuffs here: https://towardsdatascience.com/ball-detection-with-computer-vision-ai-in-sports-f9ef743e0ef1
14. This one is really awesome: https://github.com/vishaltiwari/bmvc-tennis-analytics
15. https://github.com/vishaltiwari/bmvc-tennis-analytics
How did I setup the problem?
* https://medium.com/deep-learning-turkey/google-colab-free-gpu-tutorial-e113627b9f5d
* Game:
Milestones
* Track Balls on Tennis Videos --> (x,y, frame) pos of moving ball in each frame.
* Improved detection using Focal Lens and ResNet extension of TrackNet.
* Next: build a classical filter to optimise the trajectory prediction.
* How can I reduce the time complexity of detection?
* Instead of running inference on each frame: propagate the dec
** TODO: Handle the case where we dont see the players:
** DONE: Run the detector for court detection then eliminate the wrong frames.
** TODO: Steps to perform bounce. (Test this here: )
** DONE: Tennis Court detection: https://github.com/gchlebus/tennis-court-detection
** Test algorithm to determine bounce from 2D Coordinates.
**
Report:
* https://leimao.github.io/blog/Focal-Loss-Explained/
* Think what to write in the report --> Mention advantage Focal Loss, ResNet and RetinaNet.
* Mention why Tracking steps more
Business Proposal of this project
* $60,000 or more to set up on each court [1], reconstitutes shots in 3D
https://thomas-cokelaer.info/blog/2018/02/git-how-to-remove-a-big-file-wrongly-committed/
<<<<<<< HEAD * Filtering Big branch:04b6b6e41f9dfddee954cccf73ec614d92994c18