Rancang Bangun Sistem Deteksi Jatuh (Fall Detection System) pada Penghuni Rumah berbasis Convolutional Neural Network

Abstract

Falls among household members are a serious issue with significant impacts on
physical, psychological, and economic well-being. This proposal outlines the development
of a fall detection system based on a Convolutional Neural Network (CNN) integrated into
a web application using Flask. The system aims to detect fall incidents accurately and in
real-time without requiring additional devices such as IoT sensors or wearable gadgets. The
CNN model will be trained using public datasets like the Le2i Fall Detection
Dataset and UP-Fall Detection Dataset to enhance detection sensitivity and specificity. By
adopting the YOLO (You Only Look Once) algorithm, the system can process visual data
quickly and efficiently. The web-based implementation allows household members,
caregivers, or family to monitor residents easily through a user-friendly interface. The
outcomes of this research are expected to provide a practical solution to improve household
safety, enable rapid response to incidents, and minimize severe consequences caused by
delayed intervention.