Inspiration
In our cities and towns around us, it is extremely common for there to be multi-purpose trash bins, having multiple sections for different types of trash. Mainly, the use of these trash bins are to separate the recyclable from the non-recyclable. However, human error and laziness is inevitable as research showed that nearly 1/5 of all recyclable items are placed in the wrong bin.
What it does
AutoCAN, a smart trash sorting system, is used to save time and money. With normal trash bins, people often struggle and take time to decide which bin their trash should go in; and often, this decision is proven to be wrong. Using AutoCAN's smart trash bins, people can simply throw their trash in and it will automatically be sorted. Not does this just save time for ordinary people, but also relieve the labour costs for needing to manually sort offsite.
How we built it
The entire system can be split into 3 parts:
Trash Classification Algorithm: For the classification algorithm we required datasets of trash/garbage as well as a machine learning model. The dataset was acquired from the public at Kaggle.com. We utilized Google's Teachable Machines web tool to create and train the ML model using the datasets.
Hardware: We used an Arduino servo and motor to open the trash lid and spin the trash bin respectively so that the garbage drops into the correct category. To make the 3D model of AutoCAN's smart trash bin, we 3D printed every part.
Challenges we ran into
While debugging the code we were met with a lot of problems of the code not working properly and producing the wrong results. However, we were able to catch the problems by printing debugging information into the terminal.
Another huge problem was achieving accurate and consistent image recognition of the different types of trash. However, we easily recognized that having more diverse and large amounts of datasets aided crucially. Training time and quantity also played a big factor in helping our model get better results.
Accomplishments that we're proud of
We are really proud of the 3D modelling and printing of the smart trash bin. It took a lot of hours and many redesigns and minor changes, but every small detail mattered at the end.
What we learned
Throughout this project, we have learned a variety of new things and skills. However, the most important thing that we have learned is how to use AI. Most definitely, we haven't mastered it, but we've learned its complexities and some of its operations.
What's next for AutoCAN
As of now, it is only currently in its prototype stage. If there existed next steps, then one of them would be enlarging the model into a real life sized one.
Log in or sign up for Devpost to join the conversation.