Our logo, modeled after a baby mobile and Baby's Breath flowers.
Our logo, modeled after a baby mobile and Baby's Breath flowers.
Fall 2025
PURPOSE
To make Sudden Infant Death Syndrome (SIDS) prevention affordable and accessible for all families.
SIDS - Sudden Infant Death Syndrome
RTSP - Real Time Streaming Protocols
CONTEXT & MOTIVATION
SIDS, also known as "crib death", is the leading cause of death in infants ages 1 month to 1 year old. SIDS deaths are unpredictable and usually occur when the baby re-inhales carbon dioxide in his sleep, causing it to build up in his blood. Risk factors include the following:
Placing a baby on his side or stomach to sleep, rather than on his back
Sleeping on too soft a surface, with loose blankets and bumper pads [1]
To prevent SIDS, it is possible to use a baby monitor that detects the infant's position and alerts parents if the baby turns over or begins to have trouble breathing. Solutions currently on the market cost hundreds of dollars for a specialized camera.
However, with BreatheSafe's AI image detection capabilities, families can turn the cameras that they already own into SIDS prevention devices. BreatheSafe monitors the infants breathing rate and sleeping position while they sleep, sending warnings and alerts to parents' phones as needed. If the model detects "risky objects" such as fluffy blankets or stuffed animals near the baby, the app will suggest their removal.
As long as the family's baby monitor or indoor camera uses Real Time Streaming Protocols (RTSP), as most popular models do, BreatheSafe can be integrated to help keep the baby safe while the family sleeps.
TECH STACK
We built a Python backend using MediaPipe, OpenCV, and scikit-learn to process video frames, analyze breathing patterns, and classify safe vs. unsafe sleep positions in real time.
A React Native frontend (built with Expo) connects via WebSocket, visualizing live data, breathing rate, and position confidence levels. The app also integrates OpenAI’s GPT-4 API to generate personalized safety advice and insights.
We collaborated through GitHub, built and tested components locally, and integrated everything in under 24 hours.
RESULT
BreatheSafe was built in under 24 hours by a team of 5 as a part of USC's SEP AI Hackathon, sponsored by a16z. Most of the members of our team met for the first time right before we started building. After submitting BreatheSafe for judging, we were awarded 3rd place and a $1,000 cash prize.