Inspiration
Have you ever walked up to a recycling can and wondered if you can recycle your trash. With Trash MIT, we take a picture of your item and run it through our database to check if it's recyclable.
What it does
Trash MIT identifies what an object is using a webcam and checks it against our list of items. If it is unsure of what the item is, Trash MIT asks for user input. Over time it will collect data on what is and isn't recyclable. Trash MIT has 2 purposes: (1) Collecting data on what people think is and isn't recyclable (2) telling people what is and isn't recyclable. Trash MIT could easily be implemented at restaurant trash cans where there is a small known set of trash frequently thrown away. Trash MIT makes recycling fun and interactive encouraging recycling.
How we built it
Google's cloud vision API identifies types of objects. Based on our list of recyclable objects, we then tell the user if the object is recyclable.
Challenges we ran into
We searched around online looking for a comprehensive list of recyclable items and were unable to find one. We then realized we were going to have to create our data set ourselves.
We tried using barcodes; however, it is still hard to go from identifying an object to whether it's recyclable or not.
Accomplishments that we're proud of
It runs!
What we learned
How to use API's, google cloud and OpenCV. Working in teams
What's next for trash MIT
We have so many ideas!
Trash MIT is missing hardware.
We would like to build a unit that could be attached to trash cans in urban areas. The unit would have a screen behind the trash to eliminate noise (during HackMIT we held a piece of paper behind items to stop the Google API from identifying items in the background). Instead of using 'y' and 'n' to take in user input, we would place sensors on the trash can. That way we can collect data based on what people are already throwing away. We can use this data for two purposes. Reporting back to recycling authorities on whether people are actually recycling correctly and to improve our data.
Developing the classification
We could also expand from single-stream recycling. Currently, we only identify recycling or non-recycling. We could expand so we can identify different types of recycling. We could expand to use Machine Learning to help with the identification. We could also contact local government for information on recycling laws in different areas. We could improve the interface with more color to encourage more recycling.
Log in or sign up for Devpost to join the conversation.