TerraTune

Landing page of Terra Tune
Here, user selects a polygon representing the land on which they want to plan their development
User-selected polygon is visible in red
We have several pre-defined development projects. The user will also be able to add new alternatives.
Blue boarder - land intended for development, green polygon - ecologically crucial wetland, magenta rectangles - locations for the projects

Inspiration

All around the world, local governments, planners, and decision-makers struggle to find the fragile balance between economic benefits and sustainability. When planning local investments, they should take into consideration not only the monetary returns, but also social and enviornmental concerns. The latter can be hard to quantify, thus, they are frequently overlooked. By employing AI, we can empower the decision makers to invest with the future generations in mind.

What it does

Terra Tune is a software that leverages multiple algorithms to determine the optimal use of a given plot of land. Taking into consideration returns from various investments, their ecological footprint, and the benefit for the society, it is able to provide apt recommendations on which projects to pursue. Rather than relying on conventional wisdom and intuition, local governments can now resort to data in their decision process.

How we built it

We built a website for user-friendly interaction. Thanks to Google Maps API, the user can select the plot of land that they plan to develop. Then, a linear optimization algorithm finds the optimal allocation of said land to possible projects based on multiple data sources obtained thanks to ML (we are working on adding them). Finally, genetic algorithms combined with ML models for terrain semantic segmentation (not yet added) are used to determine the most ecologically-neutral location for each of the planned developments.

Challenges we ran into

The data used by Terra Tune comes from a variety of sources. We have not yet manged to perfectly get the data and combine them with each other. Moreover, fine-tuning our models so that they reflect the actual costs and benefits to different stakeholders will be time-consuming.

Accomplishments that we're proud of

We are proud of several things. Firstly, we have fully successfully implemented the genetic algorithm for determing the location of various developments. Moreover, we have built a very easy-to-use website and successfully connected it with the backend using Flask. Finally, and most importantly, we have built a robust framework for land-use planning that can be further improved with heterogenous data sources and ML.

What we learned

We have learned that a alrge part of every project is getting and cleansing the data. Moreover, we got plenty of practice in web app development, as we had to build frontend and backend and connect it with an API. We have used a wide range of algorithms which boosted our conceptual underrstanding of those algorithms. Finally, as a team consisting of 2 people from Poland and 1 person from Pakistan, we polished our remote collaboration skills.

What's next for TerraTune

We want to increase the number of datastreams supporting our decisions, fine-tune existing models and add more AI algorithms for more automated processes. Then, we want to contact local governments in Poland, Pakistan (home countries of our team members) and abroad and start the rollout of our product.