LSTM-Human-Activity-Recognition

Why we included this project

This repo is a clear, working example of using an LSTM to classify human activities from phone sensor data. It takes the public UCI Human Activity Recognition dataset, feeds raw-ish accelerometer and gyroscope windows into a recurrent network, and distinguishes walking, sitting, standing, and other movements without hand-engineered features. The repo includes a Jupyter notebook and a standalone script, so you can follow along from downloading the data to seeing predictions. What makes it more than a code dump is the explanation: it explains what RNNs are and why LSTMs avoid the vanishing gradient problem, which is exactly the kind of context beginners need. Treat it as a teaching baseline rather than production software, but as an introduction to sequence classification, it still works well.