1Mrs.M. Anusha,2Bhargavi Mohana Maddu,3Evuri Hari Krishna,4Chadalavada Rajkumar 03 06 Elderly people and those who need constant attention are more vulnerable to falls A realtime visionbased fall detection system utilizing deep learning and human position estimate methods is presented in this study The suggested method combines MediaPipe Pose for skeletal landmark extraction from video frames with the YOLOv8 model for precise human detection Using a rulebased decision procedure with timebased validation to minimize false alarms fall events are detected by examining posture changes and body orientation The device immediately produces audible and visual alarms when a fall is confirmed A nonintrusive and effective solution for healthcare monitoring the system is developed as a Flaskbased web application that supports both live video streaming and recorded video analysis Copyright (c) Journal of Science Engineering Technology and Management Science. All rights reserved 2026