Landfill Waste Classification Using Deep Learning

Effective resource management is often hindered by traditional waste management methods such as visual inspection and manual sorting, which are subjective, labour-intensive, and not scalable. Machine learning, especially Convolutional Neural Networks (CNNs), offers a powerful alternative. Our work focused on waste disposal classification using a dataset with images of various waste types, including metal, plastic, paper, and glass. Initially, we trained a CNN, achieving high accuracy in classifying waste as biodegradable or non-biodegradable. To handle multiple complex waste objects in a single image, we used YOLOv8 for object recognition. Using the Aniopestra.v4i dataset, we achieved 95.1% mAP@50 and 76.3% mAP@50-95 in classifying materials into biodegradable and non-biodegradable categories. This approach promises to enhance waste management by providing more efficient and accurate waste classification, contributing to improved waste segregation and sustainability efforts.