Inspiration
One of the greatest challenges facing the Earth today is the exponential proliferation of trash that is taking place, at an alarming rate, as much 2.01 billion metric tons of municipal solid waste per year, according to a study by the World Bank. Since the 1960s, the amount of garbage we generate has tripled, yet most trash, recyclable or not, still ends up in the landfill. In fact, as much as 91% of the plastic we produce is not recycled, per National Geographic. If our wastes were properly sorted and recycled, experts estimate we could reduce our waste production to a quarter of what it is today. Furthermore, excess landfill waste is extremely detrimental to the environment, polluting our air and waterways, producing massive quantities of greenhouse gases, and costing communities millions in waste management services. This served as the inspiration for Garbage-Be-Gone, an intelligent and automated system to effectively sort recyclables and wastes through artificial intelligence.
What it does
In order to curb this trend, our Garbage-Be-Gone device presents a bold new step towards a cleaner, more sustainable future. By harnessing the power of computer vision and machine learning, we have devised a working prototype, which can quickly and efficiently sort a stream of waste into recyclables and trash, filtering out paper, cardboard, and plastic wastes which can be repurposed and reused. Unsorted waste is loaded into a chute where it drops individual pieces of trash onto the sorting mechanism. A camera then uses image classification to determine whether it is recyclable or not. The mechanism then tilts to the left to deposit trash and to the right to deposit recyclables.
How we built it
The image classification algorithm was trained on a Kaggle dataset using Tensorflow to classify the object as recyclable or not. We originally worked extensively with AlwaysAI's API to test and iterate on software and hardware designs. While our application was ultimately too specific for AlwaysAi's models, we were heavily inspired by their image classification and created our own similar model. Instead of a Raspberry Pi, we used a laptop to interface the computer vision model and an Arduino Uno, which controls two stepper motors. The motors drive a sorting mechanism and trash depositor, which were built with 3D-printed parts and cardboard.
Challenges we ran into
Raspberry Pi not connecting to internet, Laptop serial ports not connecting and sending data to Arduino and Raspberry Pi, finding data sets and training the model for trash classification, iterating on and designing hardware mechanisms.
Accomplishments that we're proud of
We are proud not only to have written effective software, but also create a mechanical and electrical assembly. Furthermore, we are proud of learning how to use machine learning models through AlwaysAI, as well as ultimately training and creating our own model. However, we are most proud of building our entire system out of recycled cardboard materials, repurposing waste into an effective, environmentally-friendly solution.
What we learned
Learning to use the AlwaysAI API, Raspberry Pi, and TensorFlow.
What's next for Garbage-B-Gone
In the future, we will develop and train a better model for sorting trash. We also anticipate separating based on compost and increase the space available in the trash queue.
Log in or sign up for Devpost to join the conversation.