This Human Activity Recognition (HAR) tutorial is part of the MobileHCI 2018 tutorial on using machine learning for building intelligent mobile user interfaces using the Keras deep learning library.
In this part, we will mainly cover:
01_data-preprocessing.ipynb: how to preprocess and explore timeseries data using an existing HAR dataset that uses wearable IMU sensors. We will use the USC-HAD dataset (Zhang et al., 2012)02_model-training.ipynb: how to train and evaluate a convolutional LSTM neural network model to perform supervised classification of 12 daily activities03_model-export.ipynb: how to export the trained model for use in an Android HAR app (provided there is time)
This tutorial will be held at: