Automatic Padel stroke recognition can be useful to Padel players to improve his training performances. We noticed that there is no application for monitoring an entire Padel match. We gathered raw data from a Padel match using gyroscope and accelerometer, we then investigated the stroke detection and the following classification of four different strokes. We trained a neural network from scratch, and we built an application with a smartwatch companion to collect data, moreover we used the data generated by UWB sensors together with the output from the neural network to construct a scatter plot so that the user can visualize the position of his strokes in the court.
At this link you can find the complete documentation.