wasteVision

Entire project (Software -> Arduino -> Model).
GUI
Shows the Arduino connecting to the servos.
Example of three photos as input and the 3 classifications of the input (recycling, compost, trash).

Inspiration

Most people don't take the time to sort their garbage accordingly, so many items end up in the trash rather than being recycled. This algorithm sorts through waste autonomously, taking away this responsibility from humans.

What it does

This project takes live pictures of items with a computer's webcam and then classifies them as compost, recyclable, or trash using Google's cloud vision API, openCV, and java. Then, this data is sent to an arduino which controls a garbage sorter which sorts representative items into their respective classifications. Also, the Java program creates a GUI to display the output.

How we built it

Software wise, Java was used with Google's cloud vision API, which had pre-trained machine learning models, and open CV to classify live images and send serial output. An arduino was programmed with C/C++ to respond to this serial output to control servo motors. This project was physically built from cardboard scraps, lots of tape, an arduino, servos, wires, and a breadboard.

Challenges we ran into

Communicating between the arduino and java running from a laptop. Implementing the API. Creating a physical model with limited resources. Classifying all the possible types of images. Taking multiple live images and classifying them in series.