DrunkGuard | Devpost

DrunkGuard

Inspiration

So I have this friend his name is Tanveer Salim. He parties every day and gets so drunk that he always ends up coming to his ex-girlfriends and gets into trouble. Hence, I wanted to make an application that is going to help my friend to stay away from trouble.

What it does

It uses computer vision to detect the size of your pupil if it's size is small for a sufficient period of time then it triggers the second test which takes your speech and matches against a predefined string using cosine similarity. If the two strings are not very similar then it sends a text to a white-listed friend and also calls a uber from your current location to your home. If it only passes the first test then the tool just sends a text to your white-listed friend.

How I built it

Python, TensorFlow, Keras, Twilio API, Uber API

Challenges I ran into

I had a hard time using Uber's API so I created an automation tool using selenium that does the job manually. I had a hard time implementing the algorithm to check the drunkness level. I also couldn't merge the mobile application.

Accomplishments that I'm proud of

I successfully was able to make a tool that uses speech recognition and computer vision to trigger Twilio's API and Uber's ride requesting script to send a text to your whitelisted friend. The application also requests an Uber ride from your current location to your home.

What I learned

I learned a lot about computer vision and how I can implement a real-time detection system. I also learned a lot about Kivy a python framework for mobile application development.