Lookin' for fire - Space Apps Challenge

Lookin' for fire| Automated Detection of Hazards

Awards & Nominations

Lookin' for fire has received the following awards and nominations. Way to go!

Global Nominee

Automated Detection of Hazards

Countless phenomena such as floods, fires, and algae blooms routinely impact ecosystems, economies, and human safety. Your challenge is to use satellite data to create a machine learning model that detects a specific phenomenon and build an interface that not only displays the detected phenomenon, but also layers it alongside ancillary data to help researchers and decision-makers better understand its impacts and scope.

FireFly: An app capable of detecting wildfires.

Summary

Firefly is a mobile aplication capable of detecting forest fires in the Orinoquía region in Colombia. This detection is performed using ML algorithms. The app allows its users, which are civilians and members of the fire departments of the municipalities, to report and receive alerts about fire incidents near them. Furthermore, it could be used to recognize regions with high risk.

How We Addressed This Challenge

We designed a mobile app to report and to be informed about wildfires in Colombia and we also developed a tool to identify possible fires based on satellite images. As a quick detection of a fire allows for a quick response, it would be possible to take action and decrease the impact generated by these hazards. We hope to inform in a better way about regions with a high probability of suffering a wildfire emergency.

How We Developed This Project

We chose this project because we were interested in a project related to data analysis and ML algorithms and knowing that wildfires could lead to serious environmental consequences, we thought that we could provide a tool that could accelerate the detection and response to wildfires.

Back-end:

First, we identified the variables that could be useful to train the ML algorithm and that were related to fires, such as smoke and high temperatures. Then, using the smoke data provided we trained the algorithm with Mask R CNN, a Python code that trains using datasets. On the other hand, satellite images were labeled using CVStudio and we trained the algorithm with FalconCV.

In this part we faced difficulties regarding the size of the dataset used, as we only used 310 images.

Front-end:

In the UX and UI we wanted to use a simple and elegant design to make it easier for the end user to understand and to use the interface. For the name, we chose the firefly insect because it has "fire" in its name and is often related to forest rangers. The shield in the logo represents the protection the firefighters give against forest fires, the red was used because it is the color of the fire truck engines.

We decided to use two types of users: civilians and firefighters. This was mainly to differentiate the type and quantity of information each one receives. Normal users or "civilians" can only report fires (with a confirmation prompt) and see fires in their vicinity. Firefighters have a more complete set of information, not only about the current fire they're fighting but also the historical data of the location. This was conceived as a way to give them more information for decision making and plans for awareness and prevention, additionaly, they don't have to confirm about the fire they're reporting assuring a quickly response.

How We Used Space Agency Data in This Project

We used the data of fire smoke provided in the AWS server to train the algorithm with Mask R CNN. Also, we downloaded satellite images obtained with the Earth Science NASA API to label them and to develop the training using FalconCV.