Sudden cardiac arrest kills one person every two minutes and home is where 88% of cardiac arrests occur. When elderly people are at home alone and suffer a cardiac arrest, their life is at risk

What it does

Using the Microsoft Azure Custom Vision APIs and cameras fitted in the house, we detect the pose of the persons in the room. Based on the fall of the person, we detect if the it is abrupt or normal. Based on that we send a notification to a relative who gets a notification on an app. It contains the video of the fall and options and allows him/her to take a decision on whether the fall is benign or fatal. The app has an option to call the person at home, to check on him/her or ring up emergency services

How I built it

We used Azure custom vision API for pose detection for the frames of data that is captured using the camera. We analyze the poses and conclude whether the fall qualifies as 'Fatal'. With React-Native we built an app that shows the video of the recorded fall. The video is saved in the cloud and then downloaded via url.

Challenges I ran into

Finding a proper dataset was difficult. Processing and predicting the data near real-time was tricky. The camera ends up capturing the frames much faster and getting the response from Vision API takes time. Had to use multi-threading.

Accomplishments that I'm proud of

Building a product that literally saves lives! What was most amazing was that Azure handle totally new environments in a reliable way. The application worked well after being trained on the data.

What I learned

Azure APIs.

What's next for Fall Safe

Share this project: