Every year more and more trash is thrown away, and a significant portion of that waste could be recycled or composted. We wanted to do something to reduce this unnecessary garbage.
What it does
A Cleaner Space is a web app that uses image recognition AI to determine whether an item is trash, recycling or compost.
How I built it
A Cleaner Space has a React front-end with a Phoenix back-end. I trained a resnet50 model using Tensorflow and then used the TensorflowJS library to run it in the website.
Challenges I ran into
Setting up the Tensorflow library on my computer and getting the model to work on TensorflowJS was quite challenging, but I eventually was able to get everything working.
Accomplishments that I'm proud of
I am proud of the fact that I was able to set up React, Phoenix, and Tensorflow to all work together successfully, which was quite a challenge.
What I learned
The model is actually quite big so the first time it is run on a computer it takes a while for all of the necessary files to download. After the first run, the data is cached so subsequent runs are much faster, however to avoid this potential turn off for clients, I would probably just run the model on the back-end in the future.
What's next for A Cleaner Space
Our goal is to improve the accuracy of the model, and provide more map functionality to make it as easy to use as possible.