Inspiration
One of the most devastating events for our natural environment and forests are wildfires. Greece as a country is really prone to wildfires and thousands of people die and have their properties destroyed every year. As a result, being a Greek and growing up with such problems has inspired me and my team to develop an algorithm during the 2022 AWS Hackathon which is able to identify whether there is a fire or smoke in the given input (image or video).
What it does
This project was mainly developed to identify and locate smoke or fire in the given image or video input and eventually alert any government officials in order to put the fire out before it gets out of control.
How I built it
This algorithm was written in python using the AWS SageMaker and SageMaker Studio Lab.
Challenges I ran into
It was extremely hard to optimize the algorithm and allow it to efficiently achieve a very high accuracy in detecting smoke and wildfires and distinguish between fog, cloudy sky images and small emerging fires.
Accomplishments that I'm proud of
Achieving a very high accuracy of more than 90%.
What I learned
How to implement DenseNet in order to perform this classification.
What's next for Wildfire Detector
The next steps include improving the efficiency of the model, the accuracy and providing more test cases and training data in order to allow the algorithm to easily distinguish between several scenarios where the image may show signs of smoke but not fire.
Log in or sign up for Devpost to join the conversation.