FireWatch

Inspiration

Wildfires are a growing concern in many regions around the world, with significant impact on human life, property, and the environment. Current methods of detecting and predicting wildfires rely on ground-based sensors, weather data, and human observations, which can be limited in accuracy and coverage.

What it does

The goal of this project is to develop a wildfire prediction model that uses satellite images and machine learning algorithms to improve early detection and prediction of wildfires. The model should be able to accurately identify areas at risk of wildfire and provide timely updates on fire behavior, size, and extent, to enable effective response and mitigation efforts

How we built it

We obtained these satellite photos from several government sources, then integrated them into one larger dataset, and then processed the images to make them consistent so that they could be used to train our model. Our neural network is built on the CNN technique, and we have a total dataset of around 42000 images, 12k of which were utilised to test our model. As of now, we have 96.11% accuracy. We are utilising Google and NASA's API to acquire live feeds of photographs of the selected landscapes of Earth, which we then input into our model, which predicts whether or not there is a fire at the specific location.

Challenges we ran into

1) Deployment of the website 2) Training our model(Basically ran into accuracy issues) 3) Making the API for our model info

Accomplishments that we're proud of

1) Higher accuracy of our model (96.11% as of now) 2) our HardWork 3) Cool frontend design

What we learned

1) Integrating a ml model to our website using django 2) Image Augmentation(Preprocessing our dataset to make it usable for our model) 3) we learnt versatilely in colaborating with people with different tech stacks

What's next for FireWatch

1) Making arrangements to cover almost all the wildlife areas 2) Getting real-time images 3) Letting people monitor a particular place and notifying them

Built With

• api
• django
• keras
• nextjs
• tailwindcss
• tensorflow