Dump It Like It's Hot

🌟Inspiration

Back in May 2025, one of our teammates spoke to the Mayor of Markham, who shared that even though Markham had already achieved an impressive 80% waste diversion rate, the ultimate goal was 100%. He emphasized the need to make Markham a smart, eco-friendly city, and suggested a gamified approach to encourage proper waste disposal. The goal of this approach is to show citizens that cleaning the environment voluntarily does not have to be a burden, but can instead be a rewarding activity. That's where Dump it Like it’s Hot, a scavenger hunt app for finding litter and trash, comes in! A large part of the inspiration for creating a scavenger-hunt-esque game was the 2021 Geocaching trend, and how many people wanted to find mini treasures hidden around the city. So we thought, what if we made the same thing but for cleaning the environment? Our web app transforms a serious problem into a fun and impactful solution.

♻️What it does

Introducing... Dump it Like it's Hot, Canada's first gamified way to encourage proper cleaning of the environment. Our app makes cleaning a fun, competitive, and rewarding community experience by turning litter-picking into interactive challenges with real-time leaderboards and a levelling system.

When a user spots litter on the ground, all they need to do to "collect" the litter is tap the submit button in the app. The app then records and scans the user's surroundings using their device's back camera in real time. As the app is recording the video, the backend utilizes machine learning, specifically PyTorch and COCO object detection dataset, to identify and categorize items as Garbage, Compost, or Recycling. This helps ensure that they dispose of the waste in the correct bin.

The entire disposal process is recorded. Once the user completes the video, it is automatically submitted and processed through Gemini's API, which verifies whether the user properly disposed of the litter. If the cleanup is successful, Gemini returns a confirmation. This confirmation, along with the user's ID (which is securely managed through Auth0 authentication), is then stored in our MongoDB database.

On the home and profile page, users can view the leaderboard to compare how much trash they've collected against their community. Additionally, the levelling system rewards users with "points" that count towards their "xp bar" with each successful submission, encouraging continued participation and regular cleanups. Users even have access to a dynamic, real-time map that visually tracks where waste they have collected, allowing them to see their personal impact and contribute to identifying litter hotspots in the community.

By blending gamification with real-world action, our app motivates communities to keep streets clean and healthier, all while having fun!

🔧 How we built it

Backend Dev: MongoDB, Auth0, Gemini API, COCO Dataset, PyTorch

In the backend we leveraged a variety of tools to ensure seamless integration with the frontend. PyTorch and COCO Dataset were used to help classify litter into specific categories as mentioned. Meanwhile, the Gemini API played a crucial role in verifying that users properly disposed of the litter by analyzing recorded footage. This confirmation process helps ensure the accuracy of the submissions and maintains the integrity of the gamified experience. In terms of Auth0 and MongoDB, Auth0 handled secure user authentication and identity management while MongoDB served as the primary database, securely storing submission records and data.

Frontend Dev: ReactJS, CSS, Leaflet

In the frontend, we leveraged a variety of tools to ensure an engaging user experience. ReactJS was used to build a dynamic and responsive app interface that allows users to easily navigate between pages, submit litter, and track their progress in real time. CSS was utilized to design a clean, modern, and visually appealing layout that enhances usability and maintains an intuitive experience. Leaflet powered the interactive map feature, enabling users to visualise litter collection points, monitor their personal impact, and identify community hotspots. All these incorporations ensured that anyone regardless of their background, could easily utilize the app and contribute meaningfully to the environment.

😱Challenges we ran into

One of our biggest challenges was connecting the frontend to the backend. Specifically, we struggled a lot with integrating Gemini API, as it was our first time working with this API to process and analyze video data. We needed Gemini to accurately transcribe and interpret the recorded footage, ensuring it could detect and verify whether users correctly disposed of the litter. Configuring the API to output results in a structured format that aligned with our backend logic required extensive testing and fine-tuning, especially when creating the system prompt.

Another challenge we faced was connecting the frontend and backend, since some of the initial methods and libraries we tried were deprecated. This caused unexpected errors and compatibility issues, forcing us to dig into the documentation, experiment with alternative approaches, and update our integration strategy.

😃 Accomplishments that we're proud of

We’re proud of how much we achieved on the technical side during this project. For the first time, many of our team members successfully worked with backend technologies and integrated multiple APIs, including Gemini for video verification and Auth0 for secure authentication. Learning to connect these tools to our frontend and ensuring smooth data flow between the client and server was a huge milestone for us. We also implemented machine learning using PyTorch and the COCO dataset to classify litter types accurately, something none of us had tried before.

📖What we learned

Beyond the technical growth we gained from experimenting with new tech tools, we also developed several soft skills that made a huge difference. We learned how to communicate effectively as a team under tight deadlines, collaborate through version control and CI/CD pipelines, and adapt quickly when things didn’t go as planned (which happened a lot!). Additionally, we learned how to leverage AI tools like Claude to help debug pieces of code, specifically when Auth0 and Gemini API refused to work.

🔧What's next for Dump it Like it's hot

• Convert the web app into a Flutter app and integrate Firebase
• Add voice output for visually impaired users when litter is detected
• Improve object detection by creating a custom dataset
• Expand the platform globally with new community-focused features
• Explore additional gamification elements and user engagement tools. This would include introducing new weekly challenges, user badges, and a more pixelated interface to give it a more game-like interface

Built With

• auth0
• coco
• gemini-api
• leaflet.js
• mongodb
• node.js
• pytorch
• react
• tensorflow