Litter Splitter

Inspiration

Have you ever found yourself staring at the bag of chips or coffee cup in your hand, unsure which bin to toss it in? Whether it's confusion or just being in a hurry to get the next thing done, we often end up guessing. Unfortunately, this leads to recycling contamination and more waste in landfills. That’s the problem we set out to solve. With Litter Splitter, we make trash sorting automatic, accurate, and effortless.

Our Project

Litter Splitter is a smart trash can powered by image recognition AI. It classifies the type of trash you’re throwing away as compost, recycling, or landfill, and then physically sorts it into the corresponding bin. No more second-guessing or sorting mistakes. It’s a simple idea with big impact: cleaner homes, better recycling, and less waste.

How It Works

Our system uses a webcam to capture an image of the trash as it’s placed in the input funnel. That image is analyzed by a Google Gemini-powered application, which classifies the item into one of three categories: recyclable, compostable, or landfill. The result is sent over Wi-Fi to an ESP32S microcontroller, which controls two servo motors — one to rotate the funnel over the appropriate bin, and another to release the trash at just the right moment. The trash can is divided into three sections to neatly separate each waste type.

How It’s Built

The hardware includes an ESP32S microcontroller, two SG90 servo motors, and a custom structure made from cardboard, all wired together using a breadboard and external power source. On the software side, we created a lightweight application using Google Gemini’s image recognition tool to handle object classification. This app sends sorting instructions to the ESP32S, which then actuates the servo system accordingly. The components are low-cost, accessible, and ideal for rapid prototyping.

Future Potential

LitterSplitter’s MVP lays the groundwork for much more. In the future, we see potential for fully offline classification with onboard AI models, voice guidance to teach users proper sorting habits, sensor-based detection for when the bins are full, and even analytics to track what users throw away most. With more time and support, LitterSplitter could become a valuable tool for households, schools, and public spaces looking to go greener.

Learning & Challenges

Through this project, we learned how to integrate hardware and software to solve a real-world problem. We worked hands-on with components like the ESP32S, servos, and breadboards, and gained experience in building end-to-end IoT systems. Some of our biggest challenges included establishing a stable Wi-Fi connection for the ESP32S, syncing the AI app’s output with our hardware, and tuning the servo movements to ensure accurate sorting. Each obstacle taught us valuable lessons in debugging, communication protocols, and teamwork.

Conclusion

LitterSplitter makes waste sorting easy, using AI and automation to reduce human error and environmental impact. It’s a smart trash can that doesn’t just react — it learns, helps, and empowers people to take one small but meaningful step toward a greener future.