RecyclingSorter

Inspiration

In summer 2022, as part of outreach for building recycling awareness in our community, my city sent flyers to each household about how to recycle our waste correctly. I saw it ended up in the trash after people just read them once and forgot all about them. This got me curious about the city putting in so much effort into increasing awareness and if there was a real need for spending so much tax dollars on that. I started digging and found out about recycling contamination and eventually ended up relating it to the source like ourselves. This inspired me to evaluate the reasons for improper waste sorting at the source and think of ways to combat the problem of recycling contamination. I, therefore, decided to start building the Recycling Sorter App to help people sort their waste correctly at source.

What it does

This application helps you sort recycling into three bins and tracks the weekly waste collection day so that you do not have to think too much about it.

How we built it

The part that sorts the recycling has a TensorFlow Lite model at its core. The model used was an SSD MobileNet v2 model used for image classification that we retrained on the TrashNet dataset using transfer learning and Keras. This app is an Android app that was built using Android Studio. It uses Android, AndroidX, CameraX, and TensorFlow Lite APIs.

Challenges we ran into

CameraX's permissions scheme and TensorFlow Lite's inferencing were causing a lot of trouble in the early stages of the application.

Accomplishments that we're proud of

We are very proud of creating an application that helps people with their recycling as well as reminding them about when they need to take out the recycling.

What we learned

We learned transfer learning with Keras, how to build an Android app for computer vision, and how to use cameras with Android apps.

What's next for Recycling Sorter

In the future, we plan on adding more recycling events, augmenting our dataset to build a better model, and using push notifications to inform about Trash Day.

Built With

• android
• android-studio
• androidx
• camerax
• colab
• java
• keras
• python
• tensorflow
• tensorflow-lite