In Rochfort, one of the biggest problems we face is seasonal floods and droughts. Often, these floods and droughts are devastating and can destroy crops, cause a lot of water damage, and sometimes water shortages in the town. Unfortunately, these droughts and floods are quite unpredictable and happen almost randomly throughout the year. As a result, many of us live in fear of the next flood or drought, as in the past, we've taken major financial hits due to their un-predictiveness.
What it does
H2Flow is a web app that gives officials and citizens the ability to predict exactly where and when the next flood/drought will be, and the severity of its impact. Essentially, this allows us and others to plan our moves better since we are able to track the unpredictable water cycles. We have a user-friendly front-end that gives them numerical data that is easy to understand and digest, as well as a heatmap with most likely droughts/floods in Rochfort. We believe this will save our town hundreds of thousands of dollars, as a lot is lost to water damage, or water shortages every year, especially crops. In our web app, we also have a report and resources area, where users are able to report abnormal floods or droughts in their area (in order to feed our model more data) and to let them know how they can use our data effectively to protect themselves and their assets.
How we built it
On the backend, we have a Tensorflow/Keras model that returns its prediction on when and where the next flood/drought will be. To display the results, we use Flask to serve our front-end, which shows a heatmap of flood/drought predictions, as well as easy-to-read numerical data.
Challenges we ran into
One challenge we ran into was deploying our backend onto App Engine. It would keep giving us errors every attempt, but eventually, we were able to fix the root of our problem. As well, the dataset we worked with was tough to implement in our model, as it wasn't like a conventional dataset. However, just like our other challenges, we were able to get over them.
Accomplishments that we're proud of
We are proud to have released a model that is able to accurately predict the location and time of the next flood or drought. Not only does this serve as an amazing hackathon project, but we believe it will benefit us in real life. In the future, we will continue tuning and updating our model and web app, so that we can benefit from it as much as possible. From this, we will no longer live in fear of floods and droughts in our town :^)
What we learned
We've learned a lot about front-end on the way, as none of us are experts on front-end, and usually work on backend during hackathons.
What's next for H2Flow
Due to the lack of available data, we were only able to gather data for Rochfort to train our deep learning model. However, in the future, we hope that more water data will be available, allowing us to extend our services beyond Rochfort, and all over the country, and even continent. We believe our service can make a genuine change in protecting lives, due to its accuracy (on the test data), and because of the lack of services that exist like this.