dmbee/SPAR-dataset ? reverse-engineered prompt

Build me a clean Jupyter Notebook project that loads this shoulder physiotherapy smartwatch dataset and helps classify the 7 exercises from the inertial sensor data.

I want a simple workflow that starts by loading the data from the csv files or the saved binary file, shows the structure of the signals, and does a little exploratory analysis so I can understand what is inside. Then train a basic model that can predict the exercise type from the 6 axis accelerometer and gyroscope data, and report how well it does. Please include a demo notebook that walks through the full process in a clear way, with a few visualizations and understandable output.

If useful, set it up so I can easily try different models or feature extraction approaches later. Keep it practical and easy to run from scratch, and if you need to check current docs for any libraries, feel free to look them up online.