HomeSafe

Home page

Inspiration

When tasked with producing an idea which makes Kingston a better place to live, work, and play, one of the most important factors is resident safety. We all know the feeling of sitting at home, anxiously waiting for family, friends, and loved ones to come home after a long night out. Our goal is to provide data-informed routes that people can wholeheartedly trust.

What it does

Our web app, HomeSafe cross-references historical crime data with an A* pathing algorithm through a generative AI model, producing the safest, most efficient route home possible. The HomeSafe web app also includes a safety-tailored AI chatbot to assist with any questions the user may experience.

How we built it

Initially, we wanted to take an A* algorithm, implemented through the MapBox API, and overlay crime data to aid generative AI in identifying local trends. In order to modify our routes, we inputted all crime data into our generative AI, the OpenAI API, where a combination of A*-identified shortest routes and historically safe routes produced the best route to take home.

Challenges we ran into

One of the biggest challenges facing this project was the lack of accessibly-formatted publicly available crime data, particularly in the city of Kingston. Due to this, the generative AI does not have enough data to identify longer-scale trends, such as seasonal and time-based trends.

Accomplishments that we're proud of

We are particularly proud of the integration of multiple datasets and APIs to create a single model which identifies safe routes home. As a group, we were unsure if we would be able to confidently produce a functional model which would have a lasting impact on society. Despite this, over the course of the 36-hour challenge, we were not only able to produce something that functionally works, but a project that we can and will continue to expand upon.

What we learned

Once of the aspects our team was weakest in was front-end web app development. Taking this as a challenge, our team worked tirelessly to create a web app using HTML, JavaScript, and CSS with a user interface that is both visually appealing and accessible to all.

What's next for HomeSafe

The continuation of converting publicly available historical crime data into formats which are accessible for generative AI models is something the team will continue to do in order to improve the efficiency and accuracy of the HomeSafe model. In the future, this model would be easily transferrable to any geographical location, provided that the appropriate crime data and walkway mapping is available for the region.