The current pandemic has made us aware of how important public space is and we strongly believe that public space should be...well... public. Everyone should have access to it!
Unfortunately, many urban spaces are purposefully designed to deter the most vulnerable people from using them. This is called hostile architecture, an urban-design strategy that aims to prevent homeless people from sleeping or resting.
Hostile architecture does not solve homelessness and only makes the lives of homeless people even harder. HackTheUrban.Space empowers citizens to make their community a more welcoming space for everyone!
What it does
We provide a platform for citizens to denounce hostile architecture. Users can take pictures of hostile architecture around them. We automatically check their submissions with a machine learning model and stream them using a Twitter Bot to raise awareness. Users can also see a map with all of the locations in our database.
How we built it
At Hackathons speed matters, so the right tools are essential. For the back-end we used Flask. Our database is Cassandra. The image classification model is built with FastAI.
Challenges we ran into
Twitter API authentication, Front-Back communication, Deploying the machine learning model, Training a model with limited data, Creating a dataset, Combining codebases together.
What's next for HackTheUrban.Space
We want to augment the app with a Google Street View bot that walks through cities and automatically identifies hostile architecture. Before that happens, we also need to improve the classification model as we collect more examples of hostile architecture.