Garbage Overflow Detection Algorithm Based on Improved YOLOv8n

Abstract

Unprecedented urbanization and population growth have led to an increasing amount of domestic waste, posing significant challenges to urban environments and aesthetics worldwide. This paper proposes an improved garbage detection system based on YOLOv8n. The system uses target detection algorithms to identify garbage and trash cans in road images, determining if trash cans are overflowing and need cleaning. Firstly, a Deformable Attention Mechanism (DAT) is introduced into the backbone network, incorporating dynamic sampling points. Secondly, an auxiliary head is integrated into the model's head for training, providing deep supervision. Finally, a novel Inner-MPDIoU loss function is proposed to offer more precise evaluation results and enhance generalization ability. Comparative experiments show that the algorithm achieves a mean Average Precision (mAP) of 94.0%, significantly improving feature extraction and detection accuracy.