Concept
In a future, Earth has become a waste land, devastated by unchecked waste and human irresponsibility, waste management has become humanity’s last line of defense against complete ecological collapse. This project explores a speculative design concept where individuals are held accountable for the waste they generate through hyper-surveillance and strict daily limits.
Process
The Waste Watcher Trashcan integrates Arduino, ml5, and p5.js to create a trash disposal system that tracks and records the user, their trash type, and quantity.
- Arduino
For my physical part of the project, I used a ultrasonic sensor, mini servo motor, and an RGB LED light to first build out the basic interaction for my automatic trashcan opening and closing.
Once the sensor detecs an object within 30cm, it triggers to servo motor to move/”open”. The RGB light turns green when an object is detected and changes to red when the motor moves back to original position and “closes”.
This part was pretty simple and fun to build out until I had to combine it with my physical trashcan prototype which was harder for me.
I was able to find an old cake box and turned it into my trashcan! I used two popsicle sticks to extend the handles on the motor so it can prop open the lid.
Below is an sample code that I pulled from Autodesk Instructables with my own modifications. The logic to how my automatic trashcan works.
2. Teachable Machine ml5
Using ml5 I trained 2 models, one to recognize different faces and the other one to detect the type of trash I am throwing away.
Since this is my first time working with ml5, I didn’t want to make it too complex with too many classifiers. I chose 3 main classifiers: myself (Shareen), Michael (2nd person), and no one.
I named the second person Michael because when I present in class, I would have the professor scan his face in teachable machine to upload his face data in real time.
For the second model, I only trained it to read a bottle and a crumpled paper as my two types of trash.
As you can see, the system sometimes struggle to identify the trash with full 100%. There are many factors that influences the accuracy of the detection but a main factor is because there was not enough pictures scanned of the bottle in different backgrounds into teachable machines.
3. P5js
This section for me was by far the hardest. I had to connect my p5js code with ml5 system but also connect it with my Arduino code, where my Arduino code has to read the data input from my p5js code.
Using my laptop’s camera, the system gets to scan my face from the trained ml5 model and detects what trash I scanned (can see output in console section).
The top left corner of the camera input shows a tally of how much trash you scanned compared to another person.
The center bottom of the screen shows the name of the current person in the frame that the camera is detecting.
P5js code: https://editor.p5js.org/shareenchang/sketches/mvX_4L-nr
Demo
Each individual is limited to discarding three pieces of trash per day, with their trash count made publicly visible. Should they attempt to exceed this limit, the system locks the trashcan and displays a bold, unmissable warning, signaling their failure to comply with environmental responsibility. This speculative design questions the consequences of over-surveillance and extreme measures for enforcing ecological responsibility in a world pushed to its limits.
Professor Nitsche’s try at the WasteWatcher system
Professor Nitsche’s face was scanned on the spot. But the system isn’t perfect at detecting his face yet. He was able to pose as me to throw away trash. In the future, there can also be identity theft issues of people posing as someone else to throw away trash!
Huge thanks to my peers and TA Supratim for helping with this assignment! Also, thank you to Professor Nitsche for a wonderful semester!
Leave a Reply