Automatic Bio-medical Waste Segregation Bin

Problem

Problem Statement

The project identifies that hospital waste disposal workflows rely heavily on manual sorting into color-coded bins. This introduces human error and contamination risk, and makes correction difficult once waste is disposed.

Approach

Solution Approach

Designed an automated biomedical waste bin that detects waste categories in real time and routes items to the correct compartment with minimal manual intervention.

System Architecture

• Object detection model runs on Raspberry Pi (Model 3B).
• Detection output is sent from Raspberry Pi to Arduino UNO over serial communication.
• Arduino controls the bin actuation mechanism using motor interfaces.
• Sensors are used to support object detection/positioning during disposal flow.

Hardware & Software Used

• Single Shot Multibox Detector (SSD) for object detection.
• TensorFlow 1.13.0 and OpenCV for model execution and computer vision pipeline.
• Raspberry Pi Model 3B for edge inference.
• Arduino UNO and Arduino IDE for embedded control logic.
• Motor components including stepper/servo mechanisms for physical routing.

Dataset & Model Workflow

• Custom dataset prepared for the target waste classes used in the prototype.
• Dataset split into training and testing sets (80% training split).
• Model converted and deployed on edge device for real-time inferencing.
• Classification result drives actuator commands for bin selection.

Outcome

• Validated a working end-to-end prototype from detection to physical segregation.
• Demonstrated feasibility of reducing manual sorting dependency in hospitals.
• Established a scalable base for safer, smarter biomedical waste handling workflows.