People affected by Alzheimer's disease often have difficulty remembering people's names, even their loved ones. We want to build an app that automatically scans the people around them, and identifies the people they know.
What it does
This app uses facial detection and recognition to automatically identify known faces. Text-to-speech is used to say the names of people identified.
- It lets user upload faces of people they know
- It streams from any web' or phone camera and recognizes people in real-time
- It's a cross-platform web app so it works on any mobile system (Android, iOS, Windows Phone) and desktop.
It's like facebook name tagging in with a real-time camera stream, plus text-to-speech.
How we built it
Back-end is built with Google App Engine, in Go. Front-end is built with mostly Bootstrap. Facial detection is done with tracking.js and facial recognition relies on Kairos API. Some new HTML5 tech is used, like the camera stream in browser.
- Media Capture Stream, a feature in HTML5, is not widely supported by browsers. It only works on newer versions of Chrome and Firefox.
- Tracking.js is relatively slow on mobile.
- Sometimes user can only use the front camera but not the back one.
What we learned
HTML5 is powerful! Progressive web apps could be the future of mobile