Automated identification of solid waste is critical for environmental management, yet existing object detectors often exhibit performance degradation under complex, real-world conditions. This study presents a rigorous comparative analysis of the YOLOv11 and YOLOv12 models for detecting solid waste in diverse and challenging environments. Trained on a comprehensive dataset of 4,404 images featuring variable lighting, occlusion, and background clutter, both models were evaluated on their detection accuracy and localization precision. Our results demonstrate that YOLOv12 significantly outperforms YOLOv11, achieving a mean Average Precision (mAP@0.5:0.95) of 70.73%, a 3.5% relative improvement. This enhancement is attributed to YOLOv12’s advanced architectural components, such as Area Attention (A2) and the Residual Efficient Layer Aggregation Network (R-ELAN). This study establishes YOLOv12 as a more robust and effective solution, offering a promising pathway for developing next-generation automated waste management systems.