Inspiration
I wanted to build an AI system that could protect elderly patients in nursing homes by detecting falls in real time — especially for those who may be unable to call for help themselves.
What it does
FallWatch uses a YOLOv8-based computer vision model to monitor live camera feeds, detect when a patient falls, and immediately alert nurses through a centralized dashboard that displays each patient’s status.
How we built it
I trained a fine-tuned YOLOv8 model in AWS SageMaker using the UR Fall Detection dataset, stored the dataset and model weights in S3, attempted to deploy the model on EC2, and built a web-based dashboard that allows nurses to monitor multiple patient rooms in real time.
Challenges we ran into
I faced challenges optimizing real-time inference speed on cloud instances, maintaining accuracy under varying lighting conditions, and designing an intuitive dashboard that could update dynamically without lag.
Accomplishments that we're proud of
I successfully built a full end-to-end AI pipeline—from training to deployment—that detects falls with high precision and provides an immediate visual alert to caregivers through an interactive live dashboard. This was also my first time working with AWS.
What we learned
I learned how to fine-tune pre-trained object detection models, leverage AWS cloud tools for scalable machine learning workflows, and integrate AI-driven insights into a user-friendly monitoring interface for healthcare applications.
What's next for FallWatch
I plan to expand FallWatch to support multi-room tracking, integrate mobile and voice alerts, and further refine the detection model to reduce false positives in diverse real-world environments.
Log in or sign up for Devpost to join the conversation.