sustAInabilitiy

You want to see how your city looks like without car? With a park in the middle of the city? You are not a designer or a landscape architect? We make it possible for you!

Comment

Inspiration

"Globally, there has been a steady decline in urban tree cover. Urban trees can help cities achieve 15 of the 17 UN sustainability goals" (World Economic Forum, 2022).
Green spaces provide a multitude of health, social, biodiversity, economic and of course environmental benefits (natural climate control and Carbone Dioxide vacuums).

Many communes, cities or campuses (such as EPFL) want to be more sustainable. Indeed, it can be difficult for these institutions to plan these changes as it requires the help from experts (designers and landscape architects).

What it does

Our sustAInability application detects automatically the objects (e.g. cars, people, roads) from your images via different sources (phone, webcam, file) and replaces it with sustainable objects (eg. Tree, Bush, Park, Bicycle, Playgrounds) which will make the space more pleasant for the citizens, students or employees.

How we built it

Our application is divided into 3 components:

  • Back-end with the "Stablehorde" diffusion API and "Yolo V5"-model for object detection
  • Application Interface with FastAPI
  • Front-end with Material UI

Challenges we ran into

For this hack, we relied on two different AI models (object detection and stable diffusion) with different input/output data: the tricky part was to make the two models work together nicely !

Diffusion Model is still a promising technology, and we had to fine-tune it in order to make it less experimental and more hands-on for our problematic --> The visual results are realistic !

Accomplishments that we're proud of

  1. Finishing the project !
  2. Experimenting with diffusion models
  3. Use AI for a good cause !

We took our time to find an idea (a lot of brainstorming) but we are really proud of this hack, and the impact it can have !

What we learned

Not to eat too many pizzas

Diffusion models are truly promising !

What's next for sustAInabilitiy

This hack has been applied to some concrete examples (campus, city) but the range of possibilities is broader than that (personal use).
Moving forward, we want to share our website with friends, colleagues, and they can try it at home or in their home cities !

Built With

Share this project:

Updates