Inspiration
WeatherOrNot was inspired by SDG Goal 13 (Climate Action) and SDG Goal 11 (Sustainable Cities and Communities). WeatherOrNot hopes to increased action towards climate change to prevent such disasters. In addition, WeatherOrNot hopes to influence human development in more safe and sustainable areas. We also hoped to use our multidisciplines in math and economics to calculate side-by-side statistics in this goal.
What it does
WeatherOrNot allows to view "what-if" doomsday scenarios for what could happen to your neighborhood at the current rate of climate change. Upload a picture onto our website and our AI model will project that image 10, 20, and 30 years into the future -- alongside statistics to highlight the scale of the disasters and the economic impact.
Try it out yourself! http://weather-or-not-.glitch.me/
How we built it
- Frontend in React.js
- Python REST API written in Django
- Backend AI stable diffusion inpainting
Challenges we ran into
A major challenge we ran into was hosting such a big AI model. We had to optimize our code as much as we could (even still we have ~30 second loading times) and use local GPU compute. Luckily, one of our teammates has an extremely powerful gaming laptop :).
Accomplishments that we're proud of
Using a very new technology in stable diffusion inpainting -- extremely hard to find help online! Connecting the frontend, middleware, and backend and deploying it.
What we learned
- Increased sea level and temperature changes have far-reaching human and economic impact
- Stable diffusion models are powerful and unlock immense creativity
- victory-chart is not the React.js charting library you should use
What's next for WeatherOrNot
- Visualization of other natural disasters and climate change effects (desertification, hurricanes, loss of greenery)
- Use an algorithm to determine which natural disasters would most likely affect the location of an uploaded image
- Generation of more images; for example, every 1 year instead of every 10 years
Built With
- deep-learning
- django
- machine-learning
- python
- react
- torch



Log in or sign up for Devpost to join the conversation.