Human Activity Recognition (HAR) System Demo

Overview

This repository contains a demo of a Human Activity Recognition (HAR) system using XGBoost for the CASAS Smart Home Data Set - Dataset 17. The purpose of this system is to classify and recognize the activities performed by a volunteer in a smart home environment, using sensor data. The dataset includes sensor readings and corresponding activity labels (e.g., Meal_Preparation, Wash_Dishes, etc.) collected from a smart home environment.

Demo UI

The system is designed to process and analyze the sensor data to predict and recognize human activities based on the sensor readings over time. This demo highlights how HAR can be used in real-world applications such as smart homes, health monitoring, and context-aware computing.

Dataset

Old version of the dataset is available at CASAS Smart Home Data Set. The dataset contains sensor readings and corresponding activity labels collected from a smart home environment. The activities include:

• Relax (2910)
• Meal_Preparation (1606)
• Enter_Home (431)
• Leave_Home (431)
• Sleeping (401)
• Eating (257)
• Work (171)
• Bed_to_Toilet (157)
• Wash_Dishes (65)
• Housekeeping (33)

Note: The data is not added in the repository due to its size. You can download it from the CASAS website and place the data.txt file in the aruba folder.

Requirements

To run this project by yourself, you need:

• Python 3.7+

Install the dependencies with:

pip install -r requirements.txt

Usage

Run the following command to start the Streamlit server:

streamlit run app.py