Smart Waste Classification Using Deep Learning Techniques
REST Journal on Data Analytics and Artificial Intelligence · 2025
The Smart Waste Management system leverages deep learning and computer vision techniques to provide an efficient solution for automatic waste classification, aiming to promote environmentally sustainable waste disposal practices.This project utilizes a custom dataset comprising various types of waste such as aerosol cans, aluminium food cans, cardboard boxes, plastic bottles, food waste, and more.These waste items are categorized into four distinct classes: biodegradable, non-biodegradable, trash, and hazardous.By employing the VGG16 architecture, pre-trained on ImageNet and fine-tuned using PyTorch, the model is trained to classify waste based on these categories.The system is designed to run for 20 epochs, with early stopping applied to prevent overfitting and ensure optimal performance.A robust frontend interface is built using Streamlit, enabling users to interact with the model through two options: uploading an image or capturing one via webcam.Upon submission, the model processes the input image and displays the corresponding waste classification label along with the original image.The model's performance is evaluated based on accuracy and loss for both training and test datasets, and it is further tested with real-time examples to ensurepractical applicability.