Short video put together by MLH at https://www.linkedin.com/feed/update/urn:li:activity:6523772404886495232
Sometimes it's hard to walk a few extra steps to the recycling bin. Sometimes we just don't know what to recycle. Sometimes we just forget. What isn't lazy, has a vast landscape of information, and doesn't forget? A computer. So why not outsource recycling to a cheap IoT recycling bin?
What it does
Trashy but Smart is a smart recycling bin that automatically detects objects placed in it as recyclable or not recyclable and sorts it accordingly. USA Today estimates the results of widespread recycling to save the resources equivalent to over 1.5 million park benches and 4 million barrels of oil per year, and just a 75% recycling rate would create over 1.5 million jobs. TbS aims to simplify this process with cheap, effective, and accessible recycling bins.
How I built it
We take in images of the item using a Mini Spy Camera from the MLH Hardware Lab and send it to a web server where it is forwarded to our Google Cloud machine learning model for identification. Google Cloud then responds with a classification of recyclable or non-recyclable, which is then transformed to a message to a servo controlling an arduino to move either left or right.
Challenges I ran into
We had never worked with hardware before, so it was a challenge learning how a "continuous motion" servo was different than the servos used in every tutorial online. We then had to design a trashcan that had the ability to funnel items into two different sections using only cardboard and a base trashcan while supporting an overhead camera. It was also a challenge getting a non-internet-connected Arduino to communicate with the web sever to receive servo commands, which we accomplished by using serial communication between the Arduino and a wifi-equipped sister computer.
Our lead developer was also sick, so that hurt.
Accomplishments that I'm proud of
We finished in just 10 hours and had time to sleep (and heal)!
What I learned
How to connect Arduinos to other computers with Serial to servers to Google Cloud to another server to a camera.
What's next for Trashy but Smart
Classifying more items as recyclable and non-recyclable, which would only take about 50 images of a sample item making it a very efficient process. We would also like to fabricate our own trashcan rather than recycling pizza boxes — as good as it is for the planet, it doesn't look too appealing.