An improved YOLOv11 model for garbage classification detection

With global municipal solid waste generation exceeding 2 billion tons annually and traditional manual sorting efficiency limited to under 200 items per person per hour, the development of intelligent waste classification solutions has become imperative. This study proposes an enhanced SEA-YOLOv11 algorithm that achieves breakthrough performance through three key technological innovations: (1) a Spatial and Channel Synergistic Attention (SCSA) module that integrates multisemantic spatial attention with progressive channel-wise self-attention mechanisms, significantly improving small object detection accuracy; (2) an Efficient Up-convolution Block (EUCB) employing depthwise separable convolutional structures to reduce computational costs by 23%; (3) a novel Adaptive Down-sampling (ADown) module that maintains complete feature integrity while reducing parameters by 15%. Experimental results demonstrate the model achieves 0.517 mAP50 (a 3.8% improvement) on the garbage classification dataset, with 15% fewer parameters and 11.1% lower computational requirements. This research provides an effective solution for intelligent waste classification systems.