Inspiration
When you recycle a water bottle, do you leave the cap on or off? We don't know the answer, and neither do you. If only we could stay eco-conscious consumers despite the knowledge gap.
What it does
The user places a waste product on the platform. After a few seconds, Dumpster Diver determines what kind of waste product it is (i.e. recycling, garbage), funneling the item down to the appropriate bin.
How we built it
Python for the code. Some nice, handy-dandy, ol' fashioned capentry for the hardware.
Challenges we ran into
Despite taking several images to train our model, the accuracy still left much to be desired. This is where we discovered the idea of "background augmentation (Source) -- by having a diverse set of backgrounds, your AI becomes more "accurate and robust". We fixed our model accordingly by taking several pictures with a diverse set of backgrounds; it worked much better.
Accomplishments that we're proud of
The website. Golly, the website is a thing of wonder. Automatically updating after each sort? Fantastical.
What we learned
We've never used AI at such a low level before. Cool to experience new ideas, such as the aforementioned background augmentation.
What's next for Dumpster Diver
Dumpster Diver is very easily scalable. By training the model in a more robust manner and adding more categories of waste product, we'll be worldwide in no time flat.
Log in or sign up for Devpost to join the conversation.