Quantum-inspired deep learning model for organic municipal solid waste classification toward a circular bioeconomy

Area of Science:

• Environmental Science
• Computer Science
• Quantum Computing

Background:

• Organic waste is a significant component of municipal solid waste, necessitating efficient source segregation for effective waste processing and resource recovery.
• Distinguishing between visually similar organic waste subclasses presents a challenge for conventional deep learning models.
• Accurate classification is crucial for automated waste sorting systems to enhance material recovery and sustainability.

Purpose of the Study:

• To introduce Quantum BioNet 2.0, a novel hybrid quantum-classical framework for structured organic waste classification.
• To overcome the limitations of conventional deep learning models in differentiating visually similar organic waste categories.
• To improve the efficiency and accuracy of source-level organic waste segregation for enhanced resource recovery.

Main Methods:

• Developed Quantum BioNet 2.0, integrating ResNet50 feature extraction with classical dense layers and an eight-qubit variational quantum circuit.
• Employed a two-stage hierarchical classification approach: binary classification (organic/inorganic) followed by fine-grained classification of organic waste subclasses.
• Trained and evaluated the framework on a curated dataset of 9000 waste images, comparing performance against state-of-the-art classical models.

Main Results:

• Stage 1 achieved 99.44% accuracy in binary organic/inorganic waste classification, outperforming several leading classical models.
• Stage 2 attained 98.35% accuracy and 99.96% AUC for fine-grained organic waste classification, surpassing classical baselines.
• The hybrid quantum-classical approach demonstrated measurable performance gains in fine-grained classification compared to purely classical convolutional models.

Conclusions:

• Structured hybrid quantum-classical feature fusion offers significant performance improvements for fine-grained organic waste classification.
• Quantum BioNet 2.0 effectively supports source-level waste segregation and aids automated sorting systems.
• The framework shows promise for enhancing material recovery and advancing sustainable waste management practices.