Sports TT Analysis
Very good paper
TTNet: Real-time temporal and spatial video analysis of table tennis
Notes
- Tracking of players
- Sports equipment (like balls or hockey sticks)
- Human pose estimation
- Multiple levels of detection of game-related actions
- Ball Detection from Top View
- Ball Detection from full table side view
- Player Body Segmentation
- Action Recognition
- Semantic masks of the table, humans, and scoreboard as well as the ball position
- Second stage of the ball detection is introduced.
- Event spotting - According to the table tennis rules, only ball bounces, serves, and net hits are essential to keep the score updated
- The required semantic masks of three classes (humans, table, scoreboard)
- Ball detection by the global detector were achieved in the case of the longest 9-frame sequences
More Experiments
- Segmentation Models
- Image Segmentation Keras : Implementation of Segnet, FCN, UNet, PSPNet and other models in Keras.
- Image segmentation
Keep Thinking!!!
