With the deepening of sustainable development strategies and the growing emphasis on green living, the demand for intelligent and automated waste management has significantly increased. Efficient waste classification systems not only help reduce environmental pollution but also play a pivotal role in promoting smart urban governance. The integration of machine learning and computer vision has emerged as a promising approach to address waste classification challenges, enabling real-time recognition and categorization based on visual and structural data. This study explores two complementary technical approaches: image classification using convolutional neural networks (CNNs) and structured data analysis through statistical methods. It utilizes the publicly available TrashNet dataset from Kaggle (approximately 2,500 labeled images across six waste categories) and the Garbage Classification Dataset, which includes tabular information such as waste type, weight, and material attributes. The goal is to systematically evaluate various modelssuch as SVM, Random Forest, MobileNetV2, and EfficientNetbased on accuracy, inference time, model size, and interpretability. The findings provide practical insights into selecting suitable algorithms for embedded deployment and propose optimization suggestions that enhance both performance and efficiency in real-world applications of intelligent waste sorting systems.
Jiaxin Fu (Wed,) studied this question.