CleanEarth | Devpost

Inspiration

Surveys show that over 3 quarters of people have admitted to littering in the past 5 years. Globally, 9 billion tons of litter ends up in the ocean every year. We believe that litter sitting idle induces more littering, and that is why we have created CleanEarth.

CleanEarth

What it does

CleanEarth is a platform that acts as a centralized, community-based approach to cleaning up trash in communities. You can join and create events to team up with others to go clean up garbage in a certain area. Recommended locations to cleanup are created from our mobile app. Our mobile app simulates what would realistically be a drone. Photos are taken every 5 seconds then sent to our backend to analyze for trash. If our machine learning model detects enough trash, a marker will be placed on the map as a recommended place to clean and open for someone to start an event for others to join and help.

How I built it

Front-end: We created a web application using React and Redux for our state management. We used React Native for the mobile app that periodically takes pictures to simulate what would realistically be a drone flying around scanning areas.

Back-end/Machine Learning: Our API was created using Flask and SQLite to store our data. For our litter detection, we used Tensorflow and Keras with pre-trained machine learning models to perform trash identification and classification. We created a script to download roughly 1000 Google Street View images from the San Antonio area and had the identification model scan for litter to use as the demo data.

Challenges I ran into

It was our first time creating an application that required various technologies and ran into many issues connecting all the platforms to work cohesively. Figuring out how to manage our Tensorflow and Keras versions to work with both machine learning models was difficult and required a lot of research in the deep dark web. The computing power required to host the litter detection model became an issue and required us to find a provider to fit our needs.

What's next for CleanEarth

There is a lot of potential to expand the application's use. Short-term we would like to add more features to the mobile application and encourage community involvement. We could add more to user profiles to allow users to see their own and others’ contributions to cleaning their community. Long-term we would like to see drones/vehicles taking pictures for us to scan and add to our database. Fully automating the litter detection process will make clean ups much more efficient. We would also like to expand to ocean litter detection and cleanup.