Comparative Analysis of Convolutional Neural Network Architectures for Automated Waste Classification

Abstract: Effective waste classification is a critical component of modern smart waste management systems, enabling automated sorting and recycling at scale. This paper presents a comprehensive comparative study of six convolutional neural network (CNN) architectures (VGG16, ResNet50, MobileNetV3, EfficientNet-B0, AlexNet, and a custom- designed CNN) for the task of automated waste classification into six categories: cardboard, glass, metal, paper, plastic, and trash. Each model is evaluated on a standardized dataset of 2,527 images under identical experimental conditions, including data augmentation, learning rate scheduling, and transfer learning where applicable. Performance is assessed using accuracy, precision, recall, F1-score, receiver operating characteristic (ROC) curves, and precision-recall (PR) curves. Experimental results demonstrate that the custom CNN achieves the highest classification accuracy of 95.05% with a macro-average ROC-AUC of 0.9974, outperforming established architectures. ResNet50 and MobileNetV3 follow closely at 94.07% and 93.87%, respectively, offering favorable trade-offs between accuracy and computational complexity. The analysis further includes per-class performance evaluation, confusion matrix analysis, and computational efficiency comparison, providing actionable insights for deploying waste classification systems in resource-constrained environments.

Keywords: waste classification, convolutional neural networks, transfer learning, deep learning, image classification, recycling, smart waste management