The Kernel

Inspiration

The number of accidents caused directly or indirectly by potholes is very very high especially in the case of bikers in metropolitan cities like Bangalore. Even though the BBMP claims to have been fixing roads and potholes, instead of protesting, one must realize that it is very difficult for them to cover all such roads. Thats where our idea comes in...to give them a ranked prioritised list of potholes and areas in the city which need to be fixed asap.

What it does

We have made a crowdsourced android app which a biker switches on and keeps in his pocket during his journey. The app continuously stores accelerometer and GPS readings and stores them in a Log file. The file is send to a server when the user is connected to the internet after compressing the data. On the server side, we have trained a machine learning algorithm (SVM) to distinguish between 1. Smooth riding 2. Potholes 3. Speed humps and 4. Accidents based on the accelerometer readings. This data, along with the previous history of that location is used to generate a score for the pothole. These values are plotted on a heat map of the city. This gives BBMP an easy way to target what potholes to fix first.

How we built it

We have used Location Services and Sensor Manager on the client app side to get GPS and accelerometer readings. On the server side we use Scikit learns SVM with a sliding window approach to classify between pothole, accident and smooth riding. For the visualization we used Maps API.