CoachT is an AI assistant for tennis players. It is built with an Intel Curie chip that is placed on a tennis racket. The chip has built in acceleration sensor and neural net implemented in hardware in addition to a normal Arduino board.
The machine learning is carried out in the following steps:
- Acceleration reading is collected by sampling when professional players swing shots and labelled into six groups.
- Using these data the on-board neuaral net is trained.
- Using the model trained normal players swings are compared against professional players' swing and result is outputed into one of the six groups.
- Data preprocessing is done here
- The code is written as an Arduino library in CoachT/
- The final zipped file is located at CoachT.zip
- The training data and an already trained model is found at sdcard/ to be loaded
on an SD card attached to the board.
- Training data at sdcard/coachT/data/train/
- Test data at sdcard/coachT/data/test/
- Already trained model to be used by the library at sdcard/coachT/MODEL.DAT
- Examples of how to use the library CoachT/examples/
- Load the zipped library from Arduino IDE
- And see the examples
- Documentation for using the chip found here