black-hole-diver / har-model-training

The repository containing all of the codes used for data cleaning, data processing, model training, and model deployment for 'Machine Learning Based Real-time Movement Detection of Children' BSc Thesis, 2024, ELTE IK

Human Activity Recognition Model Training

This repository contains all the codes used for data cleaning, data processing, model training, and model deployment for the thesis "Machine Learning Based Real-time Movement Detection of Children" (BSc Thesis, 2024, ELTE IK).

Dataset Used

The dataset utilized in this project is derived from a study that involved 40 children, with the goal of recognizing various activities through wearable devices. The data, structured in pickle files and accompanying CSV files for labels, provide a comprehensive basis for training our machine learning models.

Data Format

The dataset includes:

• Pickle Files: Contain raw sensor data from wearable devices, capturing movement specifics of each child involved in the study.
• CSV Files: Accompany the pickle files and contain labels for the movements. Each label corresponds to a specific activity recorded in the pickle files.

Source

The methodology and detailed description of how the data was collected can be found in the following research paper:

• Human Activity Recognition of Children with Wearable Devices Using LightGBM Machine Learning