An Optimized Faster R-CNN Framework with Golden Eagle Optimization for Garbage Detection and Classification

Urban waste management poses significant challenges due to rising population density and the increasing complexity of waste types. This study proposes an optimized Faster R-CNN model enhanced by Golden Eagle Optimization (GEO) to improve garbage detection and classification. The model addresses overfitting, class imbalance, and visual similarity between waste categories. It is trained on a self-collected, labeled dataset of 3,030 images across 10 waste types, including plastics, masks, and cans. Data augmentation is applied to enhance robustness. Experimental results demonstrate superior performance with 99.04% accuracy, 98.96% precision, and a low error rate of 0.96% compared to existing deep learning approaches.