Inspiration
When people are in a rush, they often forget to check whether or not their trash is truly recyclable or disposable. We were originally inspired to use computer vision methods to detect if something can be recycled, while integrating with hardware that would automatically sort the trash into its correct bin.
What it does
Recyclops is a smart trash sorter that determines whether an item is disposable or recyclable in real-time. When it detects the item type, an LED indicator lights up green if it is recyclable and red if it is not.
How we built it
We used Apple CreateML UI to train a model on a Kaggle dataset. We trained the model ourselves with 70% training data, 20% test, and 10% validation. Classification was performed on 6 classes (trash, plastic, paper, cardboard, metal, glass) with 74% accuracy on the test set. We attempted to use motors (connected to an arduino) to rotate a platform to throw the item into the correct bin.
Challenges we ran into
Some challenges were that we were missing many of the hardware pieces needed to transfer data between our app and hardware.
Accomplishments that we're proud of
Attempting our first hardware hack and getting the app to reach 74% accuracy for classification between multiple classes on our test set.
What we learned
We learned a great deal about how to integrate software and hardware.
What's next for Recyclops
Obtaining an arduino bluetooth piece for operating a motor to rotate the platform, to fully automate recycling/disposal.
Log in or sign up for Devpost to join the conversation.