A.I.R.

Recycling correctly is a more complex task than some might think, and we believe that and easy to use mobile application that leverages AI can help them be greener

Comment 1

Inspiration

We recently read that the vast majority of recyclable products are either not recycled or incorrectly recycled, the later causing unnecessary complexity in the recycling pipeline.

What it does

We built a mobile app that leverages the power of AI to automatically make sure that an object is recyclable or not for you, saving you the hassle of looking it up.

How we built it

We trained a Resnet-50 image classification CNN to detect what object is being captured by the camera of a mobile phone, and ran the prediction through a database of object to determine if said item is recyclable or not, and any additional procedures that must be done before recycling. We have also added a map that can show you the nearest trash can or recycle bin location using Google map API to make sure that there is no excuse for not taking care of waste.

Challenges we ran into

We had to created our own dataset because there was none that we could find. We also had little to no experience with mobile development and struggled to make a nice application that would be easy to use.

Accomplishments that we're proud of

Training a CNN on our own dataset with a good accuracy and linking it to a mobile application was nothing something any of us had done before

What we learned

We learned a lot about Pytorch and Android-studio

What's next for A.I.R

Creating a better dataset so that we can detect more objects. We think we could crowdsource a more data directly inside the app.

Share this project:

Updates