Bottle classifier

Bottle on the ground
Bottle in the hand
Bottles in a pile

Inspiration

We were inspired to work on this project after learning that collected recyclables can be sold, generating a small income and helping the local environment simultaneously. We felt that this would be a useful project for a campus, as sorted recyclables are worth more, and if the process is automated, it minimizes how much people are exposed to potential sickness.

What it does

In the current stage, it is capable of identifying plastic bottles in user-submitted images. In a later stage, it would recognize bottles in a live feed and sort them, and other types of recyclables, free of human intervention.

How we built it

We based this Image detector off of a Pretrained Neural Network, ResNet. We chose to use a pretrained net as it is not feasible to train a network from scratch in 24 hours. WE used a basic GUI to demonstrate how the system works, thought this would eventually just be replaced with an automated image system, removing the need for direct intervention.

Challenges we ran into

None of our team members have more than surface level familiarity starting the day. The vast majority of the time was spent learning about the technology.

Accomplishments that we're proud of

We are proud that we all learned much, and managed to implement basic functionality in a relatively easy-to-understand format.

What we learned

We learned how a neural network is constructed, and how to use it to classify images. We also learned how to retrain networks, though we did not have the time to implement that functionality.

What's next for Bottle classifier

Ideally, more time would be given for development. This would allow campus recycling to be deposited at a central location, and be automatically sorted it an automated, hygienic fashion.