Authors - Samarth Yogesh Jadhav, Rutuja Rajaram More, Rohit Dnyaneshwar Kokate, Aditi Dinkar Kadam, Nikhil Subhash Patankar Abstract - In urban areas, where waste production often surpasses outdated human sorting methods, efficient trash management is becoming more and more challenging. Using a Raspberry Pi and a refined YOLOv8 object detection model, we provide a real-time garbage identification solution. This hybrid technique enhances detection performance for localized waste categories, including metal, batteries, and plastic containers, by combining pre-trained YOLOv8 weights with unique fine-tuning on a domain-specific dataset. A Flask-based web application provides users with a straightforward monitoring interface. Additionally, when waste is identified, the technology automatically notifies collection staff via WhatsApp, guaranteeing prompt notifications.
