Neptune

Inspiration

Last week, while scrolling through our social medias, we instantly saw a "Breaking News" alert. It was Reuters, saying there was already a deadly wildfire in Maui. We realised how little there is to defend against wildfires; at least until it's too late.

What it does

We designed a model thats able to quickly identify wildfires, almost as soon as they start by analysing satellite imagery using CNNs. Though the wildfires are almost indiscernible by the human eye, our model is able to identify them with 100% accuracy. This would allow an almost immediate response, potentially saving thousands, to millions of lives.

However, not all areas will be experiencing a wildfire at all times. Our model is also capable of calculating the probability of a wildfire occuring, as a deduction from various factors and information obtained from the satellite imagery.

How we built it

We used mainly python. Tensorflow for the models, flask for the front-end.

Challenges we ran into

We ran into a lot of errors when making the model (for example, once, due to a really dumb error, the model was predicting everything to be a 0).

Accomplishments that we're proud of

We were able to make a functional system towards the end.

What we learned

We learnt a lot about neural networks and AI, and some of their capabilities (and very major weaknesses).

What's next for Neptune

We're extremely passionate about turning this into a major environmental app (perhaps even going beyond wildfires, checking for deforestation, floods, droughts, and everything in between). We've commonly referred to this concept as being an environmental "super-app".