Sorting recyclable materials in developed countries such as U.S., which is usually done by hand, is often expensive due to high labor costs. In addition, people often use the excuse that recycling is inconvenient and time-consuming. Another challenge is that the awareness for garbage sorting is low, and people aren't sure what’s recyclable and what’s not. As the result non-recyclables are mixed with recyclables.
"If it’s not extremely easy, people won’t do it." -- Erin Schumaker, on garbage sorting
What it does
The Box streamlines the process of garbage sorting on the input side - waste are sorted by AI without human intervention before thrown into the trash can. The user only has to show the Box the garbage and the correct bin will open automatically.
- Scan: show the webcam the piece of garbage of your choice.
- Categorize: automatically analysis the garbage and opens the corresponding lid.
In our demo, we set the Box up to accept three kinds of garbage: plastic bottles, aluminum cans, electronic waste, color-coded blue, green, and yellow respectively. All garbage that does not fit the three categories are put into the last bin.
One can also open box lids manually and/or check recent activities with our Blynk app.
How we built it
- Image processing with OpenCV
- AI enabled with Google Cloud
- Connected with Blynk
- IoT device powered by NodeMcu
Challenges we ran into
- Extremely slow WiFi at the venue
- Four servos generate too much noise on power rails that interferes with serial communication with the host.
- Multiple keywords returned from Google Cloud can be synonyms that add complexity to the decision process.
Accomplishments that we're proud of
- A smart idea that can help clean up LA city
- A fully working system that can sort garbage with high accuracy.
- Super beautiful box art.
What we learned
- How to use pre-built model provided by Google Cloud Platform to our advantage.
- Build IoT systems with Blynk
What's next for The Box: AI enabled solid waste management system
- Scale it up! The same concept and mechanism can be used on larger systems like conveyor belts for municipal waste sorting.
- Replace general-purpose AI models with purpose-trained models to improve accuracy.
- Since all the processing happens on the cloud, we can replace laptops with cheap terminals that can further reduce the cost of the system.