Under the hood
From the living room to your phone, in three quiet steps.
Everything sensitive stays at step one. Only trends, alerts and stick-figure coordinates travel outward.
Step 1 · In the home
The Edge — Raspberry Pi 5
A Pi Camera with infrared night vision feeds YOLOv8 + MediaPipe pose inference running locally. Fall and inactivity detection, and room-calibrated fall-risk zones, all happen here — the raw video never leaves this box.
Step 2 · The bridge
Zero-trust cloud connectivity
A FastAPI server over Uvicorn pushes data out through an outbound-only Cloudflare Tunnel. No open ports on the home network — the house stays invisible to port scans.
Step 3 · Wherever you are
The family dashboard
A live stick-figure stream, a 7-day activity trend chart and a weekly wellbeing summary — with WhatsApp and email alerts the moment something needs attention.
Held to a standard
Performance targets IseeHome.ai is built to hit.
These aren’t marketing numbers — they’re the benchmarks the project is being tested against.
20 FPS
Real-time frame rate
<100 ms
Inference latency
>95%
Detection accuracy
<5%
False positive rate
>95%
Fall detection accuracy
IseeHome.ai — a privacy-preserving edge monitoring framework for home-based elder care.
Interim project by Sanjay Jacob · MEng Computer Vision & AI · University of Limerick