Every time you throw trash in the recycling, you either spoil an entire bin of recyclables, or city workers and multi-million dollar machines separate the trash out for you. We want to create a much more efficient way to sort garbage that also trains people to sort correctly and provides meaningful data on sorting statistics.
Our technology uses image recognition to identify the waste and opens the lid of the correct bin. When the image recognizer does not recognize the item, it opens all bins and trusts the user to deposit it. It also records the number of times a lid has been opened to estimate what and how much is in each bin.
The statistics would have many applications. Since we display the proportion of all garbage in each bin, it will motivate people to compost and recycle more. It will also allow cities to recognize when a bin is full based on how much it has collected, allowing garbage trucks to optimize their routes. In addition, information about what items are commonly thrown into the trash would be useful to material engineers who can design recyclable versions of those products.
Future improvements include improved speed and reliability, IOTA blockchain integration, facial recognition for personalized statistics, and automatic self-learning.
How it works
- Raspberry Pi uses webcam and opencv to look for objects
- When an object is detected the pi sends the image to the server
- Server sends image to cloud image recognition services (Amazon Rekognition & Microsoft Azure) and determines which bin should be open
- Server stores information and statistics in a database
- Raspberry Pi gets response back from server and moves appropriate bin