Smart Waste Sorting System
🔗View on GitHubYOLO-based IoT waste classification system using Raspberry Pi and edge computing.
Overview
Developed a YOLOv3 + OpenCV-based waste sorting system using Raspberry Pi to classify waste into three categories with 96% accuracy. Enabled automated segregation through edge device integration with JSON-based communication between sensors and actuators.
System Architecture
Edge Computing IoT System: Raspberry Pi 4 with USB camera captures waste images at 30fps. YOLOv3 model (trained on custom waste dataset) runs inference on edge using TensorFlow Lite for optimization. Classification output (Organic/Recyclable/Hazardous) triggers servo motors via GPIO pins for segregation. SQLite database logs all classifications. MQTT broker enables communication with cloud dashboard. Sensor readings (weight, bin level) via I2C interface. Local REST API provides real-time statistics. Cloud sync via WiFi only uploads summaries to reduce bandwidth.
Setup & Implementation
Trained YOLOv3 model for waste classification. Deployed inference pipeline on Raspberry Pi equipped with camera and motors. Integrated real-time communication between modules using JSON over sockets for automated sorting based on classification output.