The Pitch

  • You are simply more likely to recycle if you have to think less to do so
  • A huge problem with the recyclying industry is that when one piece of trash is in a recycling bin, the entire bin must be thrown out
  • These two problems lead to the current failure of our recycling system and solving for this can spur true environmental progress


  • The feeling of uncertainty when you are faced with a bag of chips and an intimidating abundance of cans to decide from is why we created this app
  • Rather than being forced to make an impulsive, and often incorrect, decision, we'd like to make the right decision for you.

What it does

  • User takes a picture of a disposal good and the application displays the most likely categories of recyclable or trash along with confidence values for each category

How I built it

  • Uses a Convolutional Neural Network to classify different types of waste based on a Kaggle dataset of thousands of images
  • Hosted this web app on a Heroku server and connected this with an interactive UI

Challenges I ran into

  • Implementing Heroku and connecting our frontend to our AI backend were, without a doubt, the most difficult parts of this projects

Accomplishments that I'm proud of

  • We are proud of my neural network's ability to classify images into 6 different groups with such a high accuracy
  • We are also very proud in our Heroku server and its ability to dynamically hold and conduct our AI operations to each of our images

What I learned

  • We learned about the many different ways to maximize the accuracy of a neural network
  • We also learned how to host and upload images to a Heroku server

What's next for RecycleML

  • Integrating our software in hardware that can be installed near recyclying bins to make it extremely convinient to recycle
  • Creating a mobile app so more people can actively classify different types of waste regardless of their vicinity to a recycle bin

Made by

  • Rahul Shah & Varun Nair
Share this project: