By Alex Yoon(25'), Luke Blackwell(25'), Minjune Song(25'), Sajan Shah(25'), Steven Kam(25')
The MagicBin is an enhanced trash detetection system that hopes to collectly solve the inherent recycling problem that faces our landfills. The system blends machine learning and open source softare together with the Raspberry Pi ecosystem. Our object detection software combines OpenCV and MobileNet dataset to effectively differentiate recyleble bottles from other landfill waste. Our dataset combines over 10,000 images to effectively communicate our output along with the object's desired recyclability. Our hardware system, which includes Raspberry Pi, CannaKit, and multiple Raspberry Pi servos, can efficiently recognize the output and direct the item to its proper place in our trash bin. Furthermore, our implementation of a lottery-based system using smart contracts encourage users of the MagicBin to recycle on their own, while also reawrding them for their current contribution to the MagicBin.
- CanaKit Raspberry Pi 4 4GB Starter PRO Kit - 4GB RAM
- Arducam Lens Board OV5647 Sensor for Raspberry Pi Camera
- Adafruit 16-Channel PWM / Servo HAT for Raspberry Pi
- SparkFun 16x2 SerLCD - RGB Text
- TP-Link USB WiFi Adapter for PC
- 5V 2A (2000mA) switching power supply
- Flex Cable for Raspberry Pi Camera
- Standard servo - TowerPro SG-5010 - 5010
- Servo Extension Cable - 30cm
- OPTIX Acrylic Sheet
- Rabbit Laser Cutter
- Dremel 3D40 Printer
- CAD (Soildworks)
- Miter saw
With an increased budget and additional time, our MagicBin would consist of a newer and improved trained dataset that can detect recyclables beyond bottles. With a larger dataset, our hope is to differentiate multiple recycable materials from other waste products. This would allow our MagicBin to consist of multiple compartments for each identifiable category of recyclable materials. Furthemore, with a higher quality hardware system, our MagicBin could more quickly and efficiently sort our trash with respect to its recyclability.