ARCNet: A Novel Deep Learning Model for Robust Solid Waste Image Classification in Waste Management Systems

Accurate classification of solid waste types is essential for efficient waste management. It also supports resource recovery and environmental sustainability. Image processing-based waste classification techniques have received greater attention due to their accuracy and reliability. Existing techniques failed to capture the complex features of waste images. This work proposes Attention-Enhanced Residual Capsule Network (ARCNet), an innovative deep-learning architecture for solid waste image classification in waste management systems. ARCNet integrates Residual Attention Blocks (RABs) and Capsule Network (CapsNet) layers to improve feature extraction and capture the spatial hierarchies within waste images. In addition, a Global Context Attention Module (GCAM) is incorporated for the refined analysis of waste images. The model’s parameters are optimized using the Black Eagle Optimizer (BEO). BEO is inspired by the hunting and flight behaviors of the black eagles. Experimental results show that ARCNet outperforms traditional models in terms of accuracy, precision and recall rates. This approach supports automation in waste management and offers a reliable solution for improving sorting efficiency in recycling and disposal processes