This project focuses on developing a Machine Learning-based Human Activity Recognition (HAR) system that uses smartphone sensor data (like accelerometer and gyroscope readings) to identify a person’s physical activity in real time.
The model classifies activities such as sitting, standing, walking, running, and laying, and integrates an additional step count and calorie estimation simulation to provide useful fitness-related insights.
The system leverages smartphone sensor data collected from the UCI HAR dataset, processes it through data cleaning and feature extraction steps, and trains multiple ML models (like Logistic Regression, SVM, and Random Forest) to achieve high accuracy in recognizing user activities.
The final model is deployed using Streamlit, providing an interactive GUI where users can:
Upload or input sensor data,
View real-time activity predictions,
Simulate step count and calorie burn estimates based on predicted activities and duration.
- Create and activate a virtual environment:
python -m venv .venv .venv\Scripts\activate
- Install dependencies:
pip install -r requirements.txt
- Run in terminal:
streamlit run app.py