RefuseRadar — recycling sorting on Fire TV

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for RefuseRadar — recycling sorting on Fire TV

Inspiration

We noticed that many people struggle to understand local recycling rules, leading to contamination in waste streams. We wanted to use the ubiquity of the living room screen to create a tool that promotes sustainable living. Our goal was to make recycling accessible and visually guided for every household.

What it does

RefuseRadar allows users to hold waste items up to their Fire TV-connected camera. The application processes the image to determine if the item is recyclable, compostable, or trash. It then provides specific instructions on how to dispose of the item correctly.

How we built it

We built the backend using FastAPI and hosted it on Amazon ECS Fargate for scalable compute. For the machine learning logic, we utilized Amazon SageMaker to deploy a custom classification model. The frontend was developed in TypeScript to interface seamlessly with the Amazon Fire TV environment.

Challenges we ran into

We faced significant latency issues when transmitting high-resolution images from the Fire TV to our endpoint. We also struggled with handling various lighting conditions in typical kitchen environments. We solved these by optimizing our image compression pipeline to ensure a responsive user experience.

Accomplishments that we're proud of

We are proud of achieving near-instant inference times on the Amazon Fire TV hardware. We successfully integrated the TypeScript-based frontend with our complex AWS-side architecture. The model accuracy for common plastics and paper types exceeded our initial expectations.

What we learned

We learned a lot about optimizing computer vision models for low-power streaming device interfaces. We also gained deep insights into managing containerized applications using Amazon ECS Fargate. Understanding the nuances of Fire TV development cycles was a major highlight for our team.

What's next for RefuseRadar

We plan to implement a local database to track users' recycling impact over time. We also want to expand our model to include more complex items like hazardous electronic waste. Future updates will include gamified elements to encourage consistent participation among families.

Built With

• amazon-ecs-fargate
• amazon-fire-tv
• fastapi
• refuseradar
• sagemaker
• waste