TrashToken

Sample Trash Photo

Inspiration

We wanted a way to incentivize those who care about the environment to be the change by picking up trash. Everyday, millions of pieces of litter end up in our communities, parks, and waterways. While many people want to help, there’s often little recognition or reward for their efforts. We were inspired by the idea that technology can turn small, positive actions into a movement empowering individuals to make a visible impact and fostering a culture where cleaning up is celebrated, not overlooked.

What it does

Our project is a spin-off of the popular app SweatCoin which rewards users for walking. On the other hand, TrashToken will incentivize friendly competition by rewarding users for picking up trash. Users simply snap a photo of themselves disposing of litter in a bin, and our AI verifies the action. Each verified cleanup earns points, which in the future will be tracked on a live leaderboard. Schools, clubs, or neighborhoods can compete to see who makes the biggest difference, turning environmental stewardship into a fun, social, and rewarding experience.

How we built it

We built a Python-based image detection system using the Ultralytics YOLOv8 model. We in the future will train or fine-tuned the model to recognize both trash items and trashcans in photos. The system processes user-submitted images, runs them through the model, and checks that both a piece of trash and a bin are detected. We focused on making the detection as reliable as possible, even with limited data, and designed the code to be modular so it can be integrated into a larger app or platform in the future.

Challenges we ran into

The biggest challenge was data: there aren’t many public datasets with labeled images of trashcans and trash in real-world settings. We had to settle for a data set that just had images of trash because of time constraints. Another challenge was ensuring the model could distinguish between similar objects and work in different lighting and backgrounds. Finally, we had to make sure our detection logic was strict enough to require both trash and a bin in the same image, to prevent false positives.

Accomplishments that we're proud of

We’re proud that we have a working AI system that can automatically verify when someone is actually disposing of trash. Our model can reliably detect trash and in the future will be able to detect bins in a single photo, which is a crucial first step toward building a full incentive platform. We also made our code modular and easy to extend, so it can be used as the foundation for future features like leaderboards, rewards, or community challenges.

What we learned

This experience taught us a lot, especially since we are both first years with limited coding and developing knowledge. We learned a lot about the challenges of training AI for real-world environmental tasks, especially when data is limited. We gained experience in collecting, labeling, and augmenting data, as well as in fine-tuning and evaluating object detection models. We also learned the importance of clear, strict logic in verifying user actions, and how even a simple prototype can lay the groundwork for a much larger impact.

What's next for TrashToken

Next, we want to expand our dataset and further improve our model’s accuracy. We plan to build a simple user interface so people can easily submit photos and get instant feedback. In the future, we hope to add features like user accounts, a leaderboard, and real rewards for top contributors. Ultimately, we want TrashToken to become a tool that empowers anyone to make a visible difference in their community, one piece of trash at a time.