In our school we always notice two cans next to each other. The garbage can and the recycling can. Even with the existence of the recycling can we would always see paper and bottles and other waste that should belong in the recycling ending up in the trash.

What it does

First our AI model uses a camera to detect different types of trash and classifies them into three categories(Trash, Biodegradable and Recycling).

Then it sends the data of what is it through serial to the arduino where it proccess the input and uses it to open a can(with a stepper motor)

How we built it

AI: Yolo v5 PyTorch Python PySerial

Microcontroller: Arduino Uno 3 stepper motors breadboard Power supply

Housing: Cardboard Custom 3-d Printed parts(Motor attachments) Raw 3-d filament

Challenges we ran into

Datasets were hard to come by

Datasets were mostly innacurate and bad

Communicated through pyserial from pc to arduino was tough

Number of ports on UNO was limited

Sucking too much power

Faulty Arduino

Motor Malfunctions

Enough power to open door

Faulty lots of parts

Pin shortage(Limited pina)

Accomplishments that we're proud of

First hack for one of our members

First hardware hack for each of us

First Yolov5 hack

First time dealing with motors

What we learned

  • Serial Port Connection

  • 3-d printing

  • Debugging efficiently

  • Time Boxing

What's next for Intellibin

  • Making a model with our own dataset + perfecting the model

  • Moving the main processing to a Raspberry PI or some other lost cost capable MCU

Share this project: