Smart Bin

Forward View of Smart Bin
No Object Detected
Recyclable Object Detected
Garbage Detected

Inspiration

Landfills around our homes and in our oceans grow at an alarming rate. One way to stop this is by recycling, but many people don't know or care to know what goes in the recycling bin and what goes in the garbage.

What it does

Smart bin uses a trained AI model and webcam to identify items placed on a platform and whether they are recycling or garbage, and then moves the platform with servos to dump the item into the correct bin, recycling or waste.

How we built it

We used a simple cardboard box and servos in an Arduino circuit for the physical part of our project. For software, we trained a tensor flow model to identify common waste and recyclable items. We then used serial communication from our tensor flow model to the Arduino circuit to move the item to the correct bin.

Challenges we ran into

Our servos have low torque and use weak plastic gears so that they would get caught on the smallest wire in the bin, so we moved the circuit to the outside of the box. Additionally, our first TensorFlow model was very inaccurate, so we made many adjustments to its training settings to make it accurate.

Accomplishments that we're proud of

We are very proud of training our own AI model and getting it to connect to an Arduino circuit in real-time. We also believe this is a product that could have large real-life applications. Currently, malls such as Sherway Gardens and Yorkdale hire employees to sort between trash and recycling. Our Smart Bin can eliminate the need for labor while ensuring we recycle as much as we can.

What we learned

We learned that training your own AI object identification model isn't as easy as it seems, haha. The main thing we learned was how to use AI to control Arduino circuits in real-time, an important skill for future hackathons

What's next for Smart Bin

We hope to improve on Smart Bin with metal gear servos, a TensorFlow model that can recognize more items, sheet metal instead of cardboard, and we hope to make it modular so we can add more bins as needed, potentially including a compost bin. We also hope to create a TensorFlow model that can recognize materials and use that to tell if something is recyclable or not.