DaveWatcher

An efficient machine learning app that recognizes movement and outputs if it's a human or animal.

Comment

Inspiration

This project was inspired by a (not so friendly) house guest.

What it does

We used a camera-motion sensor and Firebase's machine learning kit to implement object recognition so that users can be notified when motion is detected, as well as be notified if the motion is a human. Our end goal is to develop a cheap and practical house security system, that can make use of the billions of retired phones world wide. A personal side goal of ours is to catch a certain minuscule house guest is still alive and is scurrying around the house.

How Our Team built it

We used the IntelliJ Idea IDE with the Android Studio plug-in to build this application and used Google's Firebase to not only keep track of users, but also incorporate the built-in machine learning kit.

Challenges We ran into

Getting the camera to perform correctly was a larger than anticipated challenge. A security camera must take frequent photos on its own, so using the default camera app is not practical. We have to use lower level Android camera libraries, while dealing with various versions of APIs (for example, a well documented but deprecated version, vs. a new but poorly documented version). Once the camera was working, we had to learn how to use Firebase's machine learning libraries. None of us had experience with machine learning before, so this was a valuable experience for all of the team.

Share this project:

Updates