Skip to content
Roshil Shrestha
All projects

Proof of concept for Tampa General Hospital

Patient Monitoring & Pose Estimation

Tracks patient movement and posture in real time with YOLOX, RTMO and ViTPose, and uses Florence to drive automated reports and alerts.

Computer vision

Overview

An AI-powered monitoring proof of concept for Tampa General Hospital that tracks patients' movements and postures in real time, so staff can receive alerts and reports are generated automatically.

The problem

Keeping track of patient positioning and mobility across different hospital environments is demanding for staff. The system had to detect patients reliably, estimate pose accurately, keep false detections low and fit into existing hospital workflows.

What I did

  • Developed the monitoring system and designed its tracking pipeline.
  • Optimised post-processing to improve efficiency and minimise false detections.
  • Led the annotation process, using Florence for prompt generation to streamline dataset creation.

How it works

  • YOLOX for patient detection and RTMO for real-time multi-person pose estimation.
  • ViTPose, a vision-transformer pose model, to improve tracking accuracy.
  • Florence, a foundation model, generating context-aware prompts for automated reporting and alerts.
  • Tracking designed to work across different hospital environments.

Outcome

  • Real-time tracking of patient movement and posture, with automated reporting and alerts for healthcare staff.