Built Worlds Hackathon 2018

In October 2018, a team consisting of myself, Ben Don (MSR, Northwestern ‘18) and Ian Davis (MSR, Northwestern ‘18) participated in the BuiltWorlds Hackathon in Chicago, IL - the goal was to solve a ‘built worlds problem’ of our choice. The problem that we chose to tackle was recycling. Our solution was a prototype of a ‘Connected Trash Can’ that included a combination of hardware (an Arduino board and a webcam) and software (machine learning using TensorFlow and a preliminary dataset). Our goal was to design a prototype that would be able to classify trash as recyclable/non-recyclable. Eventually, the design can be improved and used with multiple bins so that trash is classified correctly, and the appropriate bin opens before the trash is thrown away.

The prototype that we demonstrated during the hackathon had a webcam fitted on the inside. Once an object was thrown into the trash can, a sensor detected that the lid had closed, and a picture was automatically captured. With our initial user interface, images of the previous item of trash and the current item were present for comparison. Then a log was created with the name of the image and a label to indicate if it was recyclable or not.

Over the hackathon weekend, we built a training set of images with TensorFlow for Poets (400 labeled pictures of recyclable items, 220 pictures of trash items). Our prototype also involves a social engagement component. To build a large training dataset in the future, with labeled images, it would be beneficial to get volunteers to label images.