Inspiration
The first few minutes of a wildfire are the most critical. In adverse conditions, a difference of an hour or even 30 minutes in the deployment of the firefighting personnel can have drastic consequences in the outcome of the incident, ensuring the identifying of a wildfire quickly as a crucial task.
Unfortunately, the most exposed population to this disaster are people in isolated regions: usually in small mountain towns lacking immediate access to fire stations and police departments, making them the most vulnerable to the disaster.
Having witnessed the devastation that wildfires can cause firsthand, the cause is close to our heart. While building the product, we thought of how close friends and family members might be exposed to this tragedy, and how might help them.
What it does
FireHose works in three layers:
- Prevention: FireHose’s Machine Learning model uses very simple weather forecasts (temperature, relative humidity, wind speed and the amount of rain of the day before), to estimate the risk of wildfire for a location. On the days of high risk, FireHose will send a mobile notification to the local communities at optimal times, with advice on how to reduce their exposure to risk and asking them to stay alert to any unusual activity. The model will allow the local governments to allocate their resources more efficiently given this improved channel of communication.
- Fast collaboration. FireHose enables fast and direct communication of the villagers with the local authorities. The members of the community, who know best the area, will be able to send anonymized information about any unusual activity in the community such as illegal campfires or traces of unusual smoke. The authorities will instantaneously receive a photo with its GPS coordinates to judge if the situation requires any intervention. FireHose also allows the police department to instantaneously send an evacuation message to the members of the community.
- Fast detection. FireHose’s Computer Vision system gives drones the ability to detect the presence of fires at any stage, by identifying traces of smoke. Once a police or fire drone is deployed it will identify fires, take a picture of them, and send it along a GPS location instantaneously to the fire station, who then will take action immediately. This solution is fast, cheap, and flexible: the drones can be sent to cross-check reports of community members, or just fly in “surveillance mode” at high risk times to identify any possible fire.
How we built it
Our hack comprises of two parts, the mobile app which is used by people in the area which has a high risk of fire and the web app which can be used by the authorities (such as the police) to monitor inclement weather conditions, analyze drone footage and notify people in case of danger/evacuation.
Challenges we ran into
This is the first time we were using flutter for making an android app. While it is supposed to make life easier (one codebase vs two), it had a number of constructs which slowed us down a bit. Also since it is relatively young the ecosystem isn't super mature and many of the plugins we started off with broke down and had to be replaced.
Another issue was collecting data. Large organizations such as the NASA provide data on weather conditions and predictions, but they follow proprietary tools to access them. Finding truly free and open data was hard.
Accomplishments that we're proud of
I think we've tied together a number of aspects for this hackathon. Primarily, we've tried to focus on the challenge presented to us by IBM and whisked in BRH2019's theme into it. We made our own machine learning model that drives our decision making module. Our image recognition service is also very robust with us having tested for potential false negatives.
What we learned
The impact that forest fires have on humankind as a whole is devastating and it's integral to incorporate the community which is negatively impacted the most by making them guardians of these resources.
What's next for FireHose
It'll be interesting to explore the social/community aspects of this product. Can we leverage the people around the incident to become first responders to the scene? Can we empower them to report possible incidents that can lead to a large fires to the authorities before hand? Our hypothesis is that the community is key to the solution and we hope to serve and empower them.
Log in or sign up for Devpost to join the conversation.