08 / Real-time AI

ICU

Every second counts.

ICU gives a caregiver a timely signal when a person may have fallen or is holding a hand to their chest. The goal is to prompt a human check, with the camera feed and event timeline in the same place.

A local Flask application accepts a webcam, recording, or network stream. MediaPipe pose landmarks feed time-based rules, while server-sent events keep the caregiver dashboard updated.

PythonMediaPipeFlask

Computer vision / Local prototype

Watch the movement

A sustained low, near-horizontal body position can trigger a fall event. A hand held near the chest is reported as a visible distress gesture.

Give the signal context

Annotated frames, temporal confirmation, repeated-alert cooldowns, and a timestamped incident timeline make the detector’s output inspectable.

A person still checks

Camera angle, occlusion, and intentional lying down can affect results. A gesture does not establish a medical diagnosis. SMS, calls, accounts, and persistent history are future work.

Adapted from the project README. Explore the repository for the full story.

Have something in mind?

Let's build your next idea.

Let's talk