Transform

Inspiration

I was inspired by statistics showing that over 62% of Americans fear recycling incorrectly, with 41% opting to trash items rather than risk mistakes. This has contributed to the U.S. recycling rate being just 34%, far below the EPA's estimated 75% recyclability rate.

What it does

Transform is a quick and easy-to-use mobile app (available on both iOS and Android) that helps users recycle correctly. Users can either type in the name of an item or upload a picture of it. The app then provides detailed instructions on how to recycle the item properly—or confirms if it’s non-recyclable.

How we built it

The app’s image detection feature was developed using OpenCV for preprocessing and TensorFlow to build a machine-learning model capable of classifying recyclable items. For recyclability information, I trained another ML model in TensorFlow, using a publicly available database of recyclables. React Native and Expo were used to build the mobile app interface.

Challenges we ran into

Building the image detection functionality was the most challenging aspect. Training the AI to accurately identify objects required extensive time for model tuning and debugging. Some challenges included dataset quality, managing false positives, optimizing model performance for mobile devices, and integrating the AI smoothly into the app's workflow.

What's next for Transform

I have many exciting features planned. For example, I aim to integrate the app with grocery, takeout, and shopping apps to detect user purchases and send notifications on how to recycle those items properly.

Built With

• expo.io
• javascript
• ngrok
• opencv
• python
• react-native
• tensorflow