Towards Accurate and Efficient Waste Image Classification: A Hybrid Deep Learning and Machine Learning Approach

TL;DR

This study tackles automated waste image classification by systematically benchmarking three paradigms: classical ML with handcrafted features, deep learning (e.g., ResNet variants, EfficientNetV2S), and a hybrid DL–ML framework. The hybrid approach uses a customized ResNet50 as a deep feature extractor, applies aggressive feature selection to reduce dimensionality, and trains lightweight classifiers such as SVM and LoR, achieving near-perfect to perfect accuracy across TrashNet, Garbage Classification, and Household Garbage datasets with substantially reduced inference costs. Data integrity is improved by relabeling 43 misclassified images in the Household Garbage dataset, strengthening benchmark reliability and robustness of the hybrid method. Overall, the results show the hybrid framework delivering the best performance (up to 100% accuracy on some datasets) while dramatically reducing feature dimensionality (over 95%), enabling scalable deployment in edge and resource-constrained environments.

Abstract

Automated image-based garbage classification is a critical component of global waste management; however, systematic benchmarks that integrate Machine Learning (ML), Deep Learning (DL), and efficient hybrid solutions remain underdeveloped. This study provides a comprehensive comparison of three paradigms: (1) machine learning algorithms using handcrafted features, (2) deep learning architectures, including ResNet variants and EfficientNetV2S, and (3) a hybrid approach that utilizes deep models for feature extraction combined with classical classifiers such as Support Vector Machine and Logistic Regression to identify the most effective strategy. Experiments on three public datasets - TrashNet, Garbage Classification, and a refined Household Garbage Dataset (with 43 corrected mislabels)- demonstrate that the hybrid method consistently outperforms the others, achieving up to 100% accuracy on TrashNet and the refined Household set, and 99.87% on Garbage Classification, thereby surpassing state-of-the-art benchmarks. Furthermore, feature selection reduces feature dimensionality by over 95% without compromising accuracy, resulting in faster training and inference. This work establishes more reliable benchmarks for waste classification and introduces an efficient hybrid framework that achieves high accuracy while reducing inference cost, making it suitable for scalable deployment in resource-constrained environments.

Figures (7)

• Figure 1: Overview of proposed approach for garbage image classification
• Figure 2: Sample images from the Garbage Classification dataset
• Figure 3: Sample images from the Household Garbage dataset
• Figure 4: Sample images from the TrashNet dataset
• Figure 5: Summary of the best performance from each approach across all datasets.