BruinTrash

• 

  Taking a picture of a can to determine whether it goes into recycling or landfill!

• 

  User Interface when the user has taken a picture of the item they wish to dispose of

• 

  The user now knows the picture of the trash they took belongs in recycling!

Inspiration

All over the UCLA campus, including our rooms, there are exactly three trash cans placed together: Recycle, Compost, and Landfill. Every time we need to throw something, we are somewhat conflicted about what goes where. Do coffee cups go into recycling? Do chips bags? We often mistake them for plastic and think they should go into recycling, without realizing that they have a metal lining inside that actually interferes with the recycling process, because you can't separate the metal from plastic!

What it does

When the user takes a picture of the item that they want to dispose of, BruinTrash uses Google Cloud's Vision API to detect the waste item in the image and then uses our database on Firestore to sort the item into one of three categories: Recycle, Compost or Landfill.

How we built it

Initially, we used CocoaPods to get the Cloud Vision and Firestore API's. Then we coded to accept a picture of the waste item from the user using the camera app on their phone. We then uploaded this image to Google Cloud Vision, which identified all the objects in the image. Simultaneously, we populated a Firestore database we created with our own test data. Finally, we came up with an algorithm that compared data from Firestore to the labels identified by Cloud Vision and used it to determine how to segregate the waste. All of this was done using Swift.

Challenges and accomplishments

We had some trouble determining a streamlined way of uploading images with enough detail and quality to Google Cloud, and the most efficient way to retrieve the labels identified in the image from Google Cloud Vision. We also spent some time designing an algorithm to efficiently pick out which one of the many labels returned to us by Cloud Vision was the trash item, and which category of trash it belonged in.

What's next for BruinTrash

Next, we could use a combination of user input and Machine Learning models to increase the accuracy with which we pick what labels to focus on and segregate items. BruinTrash may also see an eWaste option in the future!

Built With

• ai
• cocoa-pods
• google-cloud
• google-firebase
• google-vision
• swift